Language conversion system and text creating system using such

Abstract

A language conversion system converts an original sentence in a source-language into a translated sentence in a target-language. The language conversion system includes a original analysis unit for analyzing the original sentence so as to obtain an intermediate structure of the original, an idea center extraction unit for extracting an idea center from the intermediate structure obtained by the original analysis unit, the idea center being a central expression of an idea to be described by a sentence, an identification unit for identifying a sentence pattern including the idea center extracted by the idea center extraction unit, an information extraction unit for extracting information required by the sentence pattern identified by the identification unit from the intermediate structure of the original sentence, a decision unit for deciding an intermediate structure of a translated sentence based on the sentence pattern and request information extracted by the information extraction unit, and a creation unit for creating a translated sentence in a surface level from the intermediate structure decided by the decision unit.

Claims

What is claimed is:

1. A language conversion system for converting an original sentence a source-language into a translated sentence in a target-language, said system comprising:

original analysis means for analyzing the original sentence so as to obtain an intermediate structure of the original;

idea center extraction means for extracting an idea center from the intermediate structure obtained by said original analysis means, said idea center being a central expression of an idea to be described by a sentence;

identification means for identifying a sentence pattern including the idea center extracted by said idea center extraction means, said identification means having a counting means for counting a number of structural elements of a target-language representation pattern, each structural element being filled with an information item in the intermediate structure of the original sentence;

information extraction means for extracting information required by the sentence pattern identified by said identification means from the intermediate structure of the original sentence;

decision means for deciding an intermediate structure of a translated sentence based on the sentence pattern and request information extracted by said information extraction means; and creation means for creating a translated sentence in a surface level from the intermediate structure decided by said decision means.

2. A language conversion system for converting an original sentence in a source-language into a translated sentence in a target-language, said system comprising:

original analysis means for analyzing the original sentence so as to obtain an intermediate structure of the original;

idea center extraction means for extracting an idea center from the intermediate structure obtained by said original analysis means, said idea center being a central expression of an idea to be described by a sentence;

identification means for identifying a sentence pattern including the idea center extracted by said idea center extraction means, wherein a degree of importance is assigned to each of structural elements of a target-language representation pattern in the sentence pattern, said identification means having calculation means for calculating a degree of importance of each of the structural elements of the target-language representation pattern, each structural element being filled with an information item in the intermediate structure of the original sentence, and means for deciding a priority order based on the degree of importance obtained by said calculation means;

information extraction means for extracting information required by the sentence pattern identified by said identification means from the intermediate structure of the original sentence;

decision means for deciding an intermediate structure of a translated sentence based on the sentence pattern and request information extracted by said information extraction means; and

creation means for creating a translated sentence in a surface level from the intermediate structure decided by said decision means.

3. A language conversion system for converting an original sentence in a source-language into a translated sentence in a target-language, said system comprising:

original analysis means for analyzing the original sentence so as to obtain an intermediate structure of the original;

idea center extraction means for extracting an idea center from the intermediate structure obtained by said original analysis means, said idea center being a central expression of an idea to be described by a sentence;

identification means for identifying a sentence pattern including the idea center extracted by said idea center extraction means;

information extraction means for extracting information required by the sentence pattern identified by said identification means from the intermediate structure of the original sentence;

decision means for deciding an intermediate structure of a translated sentence based on the sentence pattern and request information extracted by said information extraction means;

creation means for creating a translated sentence in a surface level from the intermediate structure decided by said decision means;

first storage means for storing a thesaurus dictionary in the target-language; and

second storage means for storing a translation thesaurus dictionary in which head note words of the thesaurus dictionary in said first storage means are translated in the source-language,

wherein said target-language representation pattern in the sentence pattern has means for storing a surface level in the source-language corresponding to the structural elements and means for storing a meaning marker corresponding to the structural elements, and wherein said identification means has detecting means for detecting which of the surface level of the source-language information, the meaning marker and the thesaurus word fills the structural element of the target-language representation pattern, and evaluation means for evaluating a comparison score between the target-language representation pattern and the source-language information.

4. A language conversion system for converting an original sentence in a source-language into a translated sentence in a target-language, said system comprising:

original analysis means for analyzing the original sentence so as to obtain an intermediate structure of the original;

idea center extraction means for extracting an idea center from the intermediate structure obtained by said original analysis means, said idea center being a central expression of an idea to be described by a sentence;

identification means for identifying a sentence pattern including the idea center extracted by said idea center extraction means, a target-language representation pattern having a plurality of patterns different from each other, the patterns being created from the same information, said identification means further comprising storage means for storing an identification result for each sentence, comparison means for comparing an identification result obtained by said identification means with the identification result stored in said storage means, and switching means for switching a pattern to another pattern when said comparison means obtains a result that the identification results are equal to each other;

information extraction means for extracting information required by the sentence pattern identified by said identification means from the intermediate structure of the original sentence;

decision means for deciding an intermediate structure of a translated sentence based on the sentence pattern and request information extracted by said information extraction means; and

creation means for creating a translated sentence in a surface level from the intermediate structure decided by said decision means.

5. A language conversion system for converting an original sentence in a source-language into a translated sentence in a target-language, said system comprising:

original analysis means for analyzing the original sentence so as to obtain an intermediate structure of the original;

idea center extraction means for extracting an idea center from the intermediate structure obtained by said original analysis means, said idea center being a central expression of an idea to be described by a sentence;

identification means for identifying a sentence pattern including the idea center extracted by said idea center extraction means;

information extraction means for extracting information required by the sentence pattern identified by said identification means from the intermediate structure of the original sentence;

decision means for deciding an intermediate structure of a translated sentence based on the sentence pattern and request information extracted by said information extraction means;

creation means for creating a translated sentence in a surface level from the intermediate structure decided by said decision means;

display means for displaying request information required by the sentence pattern; and

input means for inputting a command indicating that missing information of the request information displayed by said display means must be supplemented or indicating that error information of the request information must be corrected.

6. A language conversion system for converting an original sentence in a source-language into a translated sentence in a target-language, said system comprising:

original analysis means for analyzing the original sentence so as to obtain an intermediate structure of the original;

idea center extraction means for extracting an idea center from the intermediate structure obtained by said original analysis means, said idea center being a central expression of an idea to be described by a sentence;

identification means for identifying a sentence pattern including the idea center extracted by said idea center extraction means;

information extraction means for extracting information required by the sentence pattern identified by said identification means from the intermediate structure of the original sentence;

decision means for deciding an intermediate structure of a translated sentence based on the sentence pattern and request information extracted by said information extraction means; and

creation means for creating a translated sentence in a surface level from the intermediate structure decided by said decision means;

display means for displaying information which is not required by the sentence pattern; and

input means for inputting an instruction whether or not the information displayed on the display means should reflect the translated sentence.

7. A language conversion system for converting an original sentence in a source-language into a translated sentence in a target-language, said system comprising:

original analysis means for analyzing the original sentence so as to obtain an intermediate structure of the original;

idea center extraction means for extracting an idea center from the intermediate structure obtained by said original analysis means, said idea center being a central expression of an idea to be described by a sentence;

identification means for identifying a sentence pattern including the idea center extracted by said idea center extraction means;

information extraction means for extracting information required by the sentence pattern identified by said identification means from the intermediate structure of the original sentence;

decision means for deciding an intermediate structure of a translated sentence based on the sentence pattern and request information extracted by said information extraction means; and

creation means for creating a translated sentence in a surface level from the intermediate structure decided by said decision means;

sentence conversion means for obtaining an original sentence and an intermediate structure thereof from the intermediate structure of the translated sentence; and

surface level sentence creation means for creating the original sentence and a surface level sentence in the source-language based on the intermediate structure obtained by said sentence conversion means.

8. A language conversion system for converting an original sentence in a source-language into a translated sentence in a target-language, said system comprising:

original analysis means for analyzing the original sentence so as to obtain an intermediate structure of the original;

idea center extraction means for extracting an idea center from the intermediate structure obtained by said original analysis means, said idea center being a central expression of an idea to be described by a sentence;

identification means for identifying a sentence pattern including the idea center extracted by said idea center extraction means;

information extraction means for extracting information required by the sentence pattern identified by said identification means from the intermediate structure of the original sentence;

decision means for deciding an intermediate structure of translated sentence based on the sentence pattern and request information extracted by said information extraction means; and

creation means for creating a translated sentence in a surface level from the intermediate structure decided by said decision means; and

display means for displaying, as a reference sentence, a literal translation sentence of the translated sentence in the source language on a screen on which the original sentence and the translated sentence is displayed.

9. A language conversion system for converting an original sentence in a source-language into a translated sentence in a target-language, said system comprising:

original analysis means for analyzing the original sentence so as to obtain an intermediate structure of the original;

idea center extraction means for extracting an idea center from the intermediate structure obtained by said original analysis means, said idea center being a central expression of an idea to be described by a sentence;

identification means for identifying a sentence pattern including the idea center extracted by said idea center extraction means;

information extraction means for extracting information required by the sentence pattern identified by said identification means from the intermediate structure of the original sentence;

decision means for deciding an intermediate structure of a translated sentence based on the sentence pattern and request information extracted by said information extraction means; and

creation means for creating a translated sentence in a surface level from the intermediate structure decided by said decision means;

detecting means for detecting a difference between information of the original sentence and information of the translated sentence;

guidance creation means for creating guidance information in accordance with the difference detected by said detecting means; and

display means for displaying the guidance information.

10. A language conversion system for converting an original sentence in a source-language into a translated sentence in a target-language, said system comprising:

original analysis means for analyzing the original sentence so as to obtain an intermediate structure of the original;

idea center extraction means for extracting an idea center from the intermediate structure obtained by said original analysis means, said idea center being a central expression of an idea to be described by a sentence;

identification means for identifying a sentence pattern including the idea center extracted by said idea center extraction means;

information extraction means for extracting information required by the sentence pattern identified by said identification means from the intermediate structure of the original sentence;

decision means for deciding an intermediate structure of a translated sentence based on the sentence pattern and request information extracted by said information extraction means; and

creation means for creating a translated sentence in s surface level from the intermediate structure decided by said decision means;

detecting means for detecting a difference between information of the original sentence and information of the translated sentence;

retrieval means for searching for guidance information based on the information having the difference detected by said detecting means; and

display means for displaying the guidance information,

wherein additional information which can not be extracted from the intermediate structure of the original sentence required by the sentence pattern in the target-language or remainder information which is not extracted in a retrieval process of the intermediate structure of the original sentence in accordance with the request from the sentence pattern in the target-language is used as the difference detected by said detecting means.

11. A language conversion system for converting an original sentence in a source-language into a translated sentence in a target-language, said system comprising:

original analysis means for analyzing the original sentence so as to obtain an intermediate structure of the original;

idea center extraction means for extracting an idea center from the intermediate structure obtained by said original analysis means, said idea center being a central expression of an idea to be described by a sentence;

identification means for identifying a sentence pattern including the idea center extracted by said idea center extraction means;

information extraction means for extracting information required by the sentence pattern identified by said identification means from the intermediate structure of the original sentence;

decision means for deciding an intermediate structure of a translated sentence based on the sentence pattern and request information extracted by said information extraction means; and

creation means for creating a translated sentence in a surface level from the intermediate structure decided by said decision means; and

storage means for storing the intermediate structure of one or a plurality of original sentences which has been previously translated into the target-language,

wherein said storage means are treated as an object from which the idea center is to be extracted.

12. A language conversion system for converting an original sentence in a source-language into a translated sentence in a target-language,-said system comprising:

original analysis means for analyzing the original sentence so as to obtain a first intermediate structure of the original sentence;

first conversion means for converting the first intermediate structure obtained by said original analysis means into an intermediate structure of an translated sentence;

translation creation means for creating the translated sentence in the surface level from the intermediate structure of the translated sentence;

second conversion means for converting the intermediate structure of the translated sentence into a second intermediate structure in the source-language; and

surface level sentence creation means for creating a surface level sentence in the source-language from the second intermediate structure.

13. A language conversion system for converting an original sentence in a source-language into a translated sentence in a target-language, said system comprising:

original analysis means for analyzing the original sentence so as to obtain an intermediate structure of the original;

idea center extraction means for extracting an idea center from the intermediate structure obtained by said original analysis means, said idea center being a central expression of an idea to be described by a sentence;

identification means for identifying a sentence pattern including the idea center extracted by said idea center extraction means;

information extraction means for extracting information required by the sentence pattern identified by said identification means from the intermediate structure of the original sentence;

decision means for deciding an intermediate structure of a translated sentence based on the sentence pattern and request information extracted by said information extraction means; and

creation means for creating a translated sentence in a surface level from the intermediate structure decided by said decision means;

detecting means for detecting a difference between information of the original sentence and information of the translation;

retrieval means for searching for guidance information based on the information having the difference detected by said detecting means; and

display means for displaying the guidance information.

14. A language conversion system for converting an original sentence in a source-language into a translated sentence in a target-language, said system comprising:

original analysis means for analyzing the original sentence so as to obtain an intermediate structure of the original;

idea center extraction means for extracting an idea center from the intermediate structure obtained by said original analysis means, said idea center being a central expression of an idea to be described by a sentence;

identification means for identifying a sentence pattern including the idea center extracted by said idea center extraction means;

information extraction means for extracting information required by the sentence pattern identified by said identification means from the intermediate structure of the original sentence;

decision means for deciding an intermediate structure of a translated sentence based on the sentence pattern and request information extracted by said information extraction means;

creation means for creating a translated sentence in a surface level from the intermediate structure decided by said decision means; and

means for selecting a target-language phrase pattern without relation to the idea center and for creating the translated sentence in the target-language using the target-language phrase pattern in a case where a structural element corresponding to an optional element in the target-language phrase pattern driven by the idea center does not exist in the intermediate structure in the source-language.

15. A language converting system applied to a machine translation apparatus having input means for inputting an original sentence, translation means for translating the original sentence into a target-language, storage means for storing information required in said translation means, and output means for outputting a result obtained by said translation means, said translation means comprising:

original analysis means for analyzing the original sentence so as to obtain an intermediate structure of the original sentence;

idea center extraction means for extracting an idea center from the intermediate structure obtained by said original analysis means, said idea center being central expression of an idea to be described by a sentence;

identification means for identifying a sentence pattern including the idea center extracted by said idea center extraction means;

information extraction means for extracting information required by the sentence pattern identified by said identification means from the intermediate structure of the original sentence;

decision means for deciding an intermediate structure of a translated sentence based on the sentence pattern and request information extracted by said information extraction means; and creation means for creating a translated sentence in a surface level from the intermediate structure of the translated sentence, and said language converting means further comprising:

confirmation means for confirming the result obtained by said translation means.

16. A language conversion system for converting an original sentence in a source-language into a translated sentence in a target-language, said system comprising:

idea center extraction means for extracting an idea center from the original sentence, said idea center being a central expression of an idea to be described by a sentence;

identification means for identifying a representation pattern including the idea center extracted by said idea center extraction means;

information extraction means for extracting information required by the representation pattern identified by said identification means from the intermediate structure of the original sentence; and

display means for displaying the information extracted by said information extraction means, said display means displaying the information required by the representation pattern in a table format, values displayed in the table format being able to edit.

17. The system as claimed in claim 16, further comprising:

output means for outputting a target-language sentence based on the representing pattern in the target-language and an original sentence based on a representation pattern in the source language corresponding to the representing pattern in the target-language.

18. The system as claimed in claim 16, further comprising:

output means for outputting a target-language sentence created based on the representation pattern with reference to the information obtained by said information extraction means and the editing result of the values displayed in the table format; and

editing means for editing the target-language sentence output by said output means,

wherein if a value changed by said editing means is a word in the source-language, the changed valued is converted into an expression in the target-language.

19. A language conversion system for converting an original sentence in a source-language into a translated sentence in a target-language, said system comprising:

idea center input means for inputting an idea center directed to input to said system by a user, said idea center being a central expression of an idea to be described by a sentence;

identification means for identifying a representation pattern including the idea center input by said idea center input means;

information extraction means for extracting information required by the representation pattern identified by said identification means from the intermediate structure of the original sentence;

display means for displaying the information extracted by said information extraction means.

Description

This application is related to prior U.S. application Ser. No. 07/626,444 filed Dec. 12, 1990 now U.S. Pat. No. 5,303,150; and Ser. No. 07/899,269 filed Jun. 16, 1992 now U.S. Pat. No. 5,338,976 for which all appropriate benefits under 35 U.S.C. .sctn.120 are claimed.

BACKGROUND OF THE INVENTION

(1) Filed of the Invention

The present invention generally relates to a language conversion system, and more particularly to a language conversion system capable of being used for the document preparation, the natural language processing, and the translation between different languages. The language conversion system according to the present invention is applicable, for example, to machine translation systems, document preparation systems, cai (Computer Assisted Instruction) systems, foreign language preparation supporting system, and other OA (Office Automation) devices such as word processors and personal computers.

(2) Description of the Related Art

FIG. 1A shows a conventional machine translation method. This method is generally referred to as structure transfer method. In this method, first, an original is analyzed, and an intermediate structure in a source-language (e.g. Japanese) which structure can be easily transferred to an intermediate structure in a target-language (e.g. English) is generated based on the analyzed results. The intermediate structure in the source-language is then transferred to that in the target-language. The translation in the surface-level of the target-language is generated from the intermediate structure therein. However, even if the logically precise translation is created in accordance with the above method, it is apt to be difficult for natives of the target-language to understand the translation between languages, such as Japanese and English, with the different cultural backgrounds.

FIG. 1B shows another conventional machine translation method. In this method, a structure independent from any languages is generated based on the analyzing results obtained by the analysis of the original in the source-language. The translation in the surface-level of the target-language is then generated from the structure. However, the difference between the cultural backgrounds of the source-language and the target-language is not considered in the above method. Thus, in this case also, it is apt to be difficult for natives of the target-language to understand the translation.

The following prior arts have been known.

1) Japanese Laid Open Patent Application No. 63-44276 "AUTOMATIC GENERATOR FOR GENERATED SYNTAX":

In this system, an idea structure is obtained by the analysis of the original in the source-language, as shown in FIG. 1B, and the surface-level of the target-language is then generated from the idea structure. This system obtains the conceptual structure which is an intermediate structure independent from any languages. However, the language structure such as the intermediate structure independent from any languages can not be obtained in actuality. As a result, the conceptual structure obtained by this system may depend on the source-language. Thus, it may be difficult for natives to understand the translation generated therefrom.

2) Kumano et al., "User-Cooperative Japanese Sentence Generation System," National Conference of the 42nd Information Processing Society (First term of 1991):

This system has bene proposed on the assumption that a user of the translation carries out operations. This is, if a translation having contents different from those of the original is obtained, the user supplies information of changing and correction and supplemental information to the system. As a result, the translation having the high quality can be obtained. Hence, the target-language knowledge is not a little required for the user.

3) Saito and Tomita, November 1984, "Automatically Writing a Letter in a Foreign Language," Symposium of the Natural Language Processing Technology:

This system has been proposed for a user having no knowledge of the target-language, in which system a letter is written in the target-language using only the knowledge of the source-language. According to this system, the letter having the high quality can be written. However, as example of sentences each of which examples has blanks, are used and required words are filled in the blanks, sentences other than the examples can not be translated.

4) Nomiyama, "Lexical Selection Mechanism in Machine Translation Using Target Language Knowledge," National Conference of the 42nd Information Processing Society (First term of 1991):

As relationships between verbs and nouns cooperated with each other depend on languages, it is hard for users to read translation. In this translation technique, suitable terms for translation are selected with consideration to this matter. However, a problem is not yet solved in that it is difficult to understand the expression of the whole translated sentence.

5) Satoshi Sato and Makoto Nagao, "Toward Memory-Based Translation":

A large number of examples of translated sentences which can be relatively easily understood are gathered along with originals corresponding to the examples, and the examples are stored in a system. A sentence closest to an original input is selected from the examples of translated sentences. Parts of the selected sentence different from those of the original are detected and corrected. According to this system, a translation having a relatively high quality can be obtained. However, it is impossible to gather examples for all sentence patterns. Thus, if there is no example corresponding to an original input, the translation can not be obtained at all.

In addition, there has been proposed a system in which contents of a source-language sentence is explained by the target-language using an idea center and target-language sentence patterns driven by the idea center so that a translation is obtained, the idea center (hereinafter, for the sake of simplicity, referred to as an IC) being a central term of an idea to be explained by a sentence. Each of the target-language sentence patterns is formed of elements essential for (close to) the IC, the elements essential for the IC being obtained by analysis of actual target-language sentences.

The popular conventional system converts the analysis (tree structure) of the source-language into the tree structure of the target-language in parts, and generates the target-language. According to this system, even if the lack of knowledge is compensated for by the interaction with the user, the system tends to take a source-language-based structure. As a result, the sentence form is unnatural for the target-language, and the expression in unnatural.

On the other hand, in the system which obtains all the information through inquiries with the user, it is necessary to prepare an extremely large amount of data in order to cope with a wide range of documents. In addition, the operation is machine initiated from the beginning and makes the user feel under control of the machine.

The information required in the target-language is very often missing in the sentence of the source-language. For this reason, the system in which there is no interaction with the user tends to frequently generate incomplete output.

Moreover, the information necessary in the target-language may not exist in the original sentence. According to the convention system, the translation is in many cases completed by supplementing such necessary information by a default process. However, erroneous information may be added and necessary information may be dropped if the default process is insufficient, and this method tends to generate an unnatural translation.

On the other hand, according to the translation system which uses the examples of translated sentences, an extremely large amount of examples of translated sentences must be prepared in order to cope with a wide range of documents.

SUMMARY OF THE INVENTION

Accordingly, a general object of the present invention is to provide a novel and useful language conversion system in which the disadvantages of the aforementioned prior art are eliminated.

A more specific object of the present invention is to provide a language conversion system in which, even if cultural backgrounds of the source-language and the target-language differ from each other, a sentence in the source-language can be precisely translated in the target-language.

Another object of the present invention is to provide a language conversion system capable of being flexibly used by a user.

Another object of the present invention is to provide a language conversion system in which a precisely translated sentence can be rapidly obtained.

Another object of the present invention is to provide a language conversion system in which a sentence having natural expressions in the target-language is obtained.

The above objects of the present invention are achieved by a language conversion system for converting an original sentence in a source-language into a translated sentence in a target-language, the system comprising: original analysis means for analyzing the original sentence so as to obtain an intermediate structure of the original; idea center extraction means for extracting an idea center from the intermediate structure obtained by the original analysis means, the idea center being a central expression of an idea to be described by a sentence; identification means for identifying a sentence pattern including the idea center extracted by the idea center extraction means; information extraction means for extracting information required by the sentence pattern identified by the identification means from the intermediate structure of the original sentence; decision means for deciding an intermediate structure of a translated sentence based on the sentence pattern and request information extracted by the information extraction means; and

creation means for creating a translated sentence in a surface level from the intermediate structure decided by the decision means.

The above objects of the present invention are also achieved by a language conversion system for converting an original sentence in a source-language into a translated sentence in a target-language, the system comprising: original analysis means for analyzing the original sentence so as to obtain a first intermediate structure of the original sentence; first conversion means for converting the first intermediate structure obtained by the original analysis means into an intermediate structure of a translated sentence; translation generation means for creating the translated sentence in the surface level from the intermediate structure of the translated sentence; second conversion means for converting the intermediate structure of the translated sentence into a second intermediate structure into the source-language; and surface level sentence creation means for creating a surface level sentence in the source-language from the second intermediate structure.

The above objects of the present invention are also achieved by a language conversion system applied to a machine translation apparatus having input means for inputting an original sentence, translation means for translating the original sentence into a target-language, storage means for storing information required in the translation means, and output means for outputting a result obtained by the translation means, the translation means comprising: original analysis means for analyzing the original sentence so as to obtained an intermediate structure of the original sentence; idea center extraction means for extracting an idea center from the intermediate structure obtained by the original analysis means, the idea center being central expression of an idea to be described by a sentence; identification means for identifying a sentence pattern including the idea center extracted by the idea center extraction means; information extraction means for extracting information required by the sentence pattern identified by the identification means from the intermediate structure of the original sentence; decision means for deciding an intermediate structure of a translated sentence based on the sentence pattern and request information extracted by the information extraction means; and creation means for creating a translated sentence in a surface level from the intermediate structure of the translated sentence, and the language converting means further comprising: confirmation means for confirming the result obtained by the translation means.

The above objects of the present invention are also achieved by a language conversion system for converting an original sentence in a source-language into a translated sentence in a target-language, the system comprising: idea center extraction means for extracting an idea center from the original sentence, the idea center being a central expression of an idea to be described by a sentence; identification means of identifying a representation pattern including the idea center extracted by the idea center extraction means; information extraction means for extracting information required by the representation pattern identified by the identification means from the intermediate structure of the original sentence; display means for displaying the information extracted by the information extraction means.

The above objects of the present invention are also achieved by a language conversion system for converting an original sentence in a source-language into a translated sentence in a target-language, the system comprising: idea center input means for inputting an idea center directed to input to the system by a user, the idea center being a central expression of an idea to be described by a sentence; identification means for identifying a representation pattern including the idea center input by the idea center input means; information extraction means for extracting information required by the representation pattern identified by the identification means from the intermediate structure of the original sentence; display means for displaying the information extracted by the information extraction means.

The above objects of the present invention are also achieved by a language conversion system for converting an original sentence in a source-language into a sentence in a target-language by use of a computer, the system comprising: a database for storing representation patterns in the target-language; evaluation means for evaluating an adaptivity factor of each representation pattern to the original sentence; retrieval identification means for searching for an identifying information required by a target-language pattern in the original sentence in the source-language pattern in the original sentence in the source-language to create the sentence in the target-language; interactive means for obtaining information required to create the sentence in the target-language in an interaction with a user; selecting means for selecting a representation pattern in the target-language adaptable to the original sentence based on the adaptivity factor obtained by the evaluation means; and output means for outputting the sentence in the target-language created based on the information obtained by the retrieval identification means and the interactive means, wherein, in a case where there are a plurality of representation patterns for representing similar matters, judging information based on which the representation pattern is selected is assigned to each of the representation patterns stored in the database, the judging information being displayed along with the plurality of representation patterns.

Another object of the present invention is to provide a text creating system in which the structure of an original text is analyzed, a typical text pattern is selected based on the analysis result, and a text in the target-language is created using the selected typical text pattern.

Another object of the present invention is to provide a text creating system in which the created sentence can be flexibly edited by a user.

Another object of the present invention is to provide a text creating system in which the structure of the sentence can be selected based on matters that the user wishes to express.

Another object of the present invention is to provide a text creating system in which a single text structure can be decided from among a plurality of text structure adaptive to the text structure of the original text.

Another object of the present invention is to provide a text creating system in which the text can be edited in accordance with the degree of importance of the text elements.

Another object of the present invention is to provide a text creating system capable of creating a text which may be various expressions.

The above objects of the present invention are achieved by a text creating system comprising: text input means for inputting a text in a source-language; a first structural element database for storing a plurality of structural elements forming texts in the source-language; original text analyzing means for detecting structural elements equal to those in the first structural database from the text in the source-language and for analyzing a structure of the text in the source-language; a first typical text structures in a target-language; first selecting means for evaluating an adaptivity factor of each of the typical text structures to the structure of the text in the source-language obtained by the original text analyzing means and for selecting a typical text structure having a highest adaptivity factor; text creating means for creating a target-language text based on the typical text structure selected by the first selecting means; and editing means for editing the target-language text created by the text creating means.

The above objects of the present invention are also achieved by a text creating system comprising: a text element database for storing text elements forming a text; a text structure database for storing information of text structures each of which is a set of text elements; a sentence pattern database for storing information used to generate sentences corresponding to each of the text elements; text element selecting means for selecting text elements from the text element database; text structure retrieval means for searching the text structure database for a text structure based on the text elements selected by the text element selecting means; text generating means for generating a target text based on the text structure obtained by the text structure retrieval means using the information stored in the sentence pattern database; and editing means for editing the target text created by the text creating means.

The above objects of the present invention are also achieved by a text creating system comprising: original text analyzing means for detecting text elements in an original text; text element database for storing text elements forming texts; text structure database for storing a plurality of text structures of target texts each of which structures is a set of text elements; sentence pattern database for storing information used to generate sentences corresponding to each of the text elements; target-text structure deciding means for deciding a text structure adaptive to a set of the text elements of the original text from the text structure database; text generating means for generating a target text based on the text structure decided by the target-text structure deciding means using the information of the sentence pattern database; and editing means for editing the target text created by the text generating means, wherein the text structure database stores degree of importance information of each of the text elements.

Additional objects, features and advantages of the present invention will become apparent from the following detailed description when read in conjunction with the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1A is a diagram illustrating a conventional translation method.

FIG. 1B is a diagram illustrating another conventional translation method.

FIG. 2 is an outside view illustrating a language conversion system according to an embodiment of the present invention.

FIG. 3 is a block diagram illustrating the structure of the language conversion system according to the embodiment of the present invention.

FIG. 4 is a diagram illustrating a structure of a translating body shown in FIG. 3.

FIG. 5 is a diagram illustrating an example of a field classified keyword list database.

FIG. 6 is a diagram illustrating an example of a field classified IC candidate list database.

FIGS. 7, 8, 9, 10, and 11 are diagrams illustrating examples of a target-language phrase pattern.

FIG. 12 is a diagram illustrating an example of a target-language phrase pattern database.

FIG. 13 is a diagram illustrating an example of a field classified translation dictionary database.

FIG. 14 is a diagram illustrating an example of a guidance database.

FIG. 15 is a diagram illustrating an example of contents of a translation state memory.

FIG. 16 is a diagram illustrating examples of an original sentence (J).

FIG. 17 is a diagram illustrating an example of contents of a extraction IC buffer.

FIG. 18 is a diagram illustrating an example of a source-language intermediate structure (S).

FIGS. 19 and 20 are diagrams illustrating examples of contents of request information buffer.

FIG. 21 is a diagram illustrating an example of contents of a guidance buffer.

FIG. 22 is a diagram illustrating an example of contents of a display buffer.

FIG. 23 is a diagram illustrating an example of a target-language intermediate structure (T).

FIG. 24 is a diagram illustrating an example of an intermediate structure (SI) of a reference sentence.

FIG. 25 is a diagram illustrating examples of a translated sentence (E).

FIG. 26 is a diagram illustrating examples of a reference sentence (J').

FIGS. 27, 28, 29 and 30 are diagrams illustrating examples of the interaction with a user through a screen.

FIG. 31 is a diagram illustrating a translation system according to another embodiment of the present invention.

FIG. 32 is a diagram illustrating a translation system according to another embodiment of the present invention.

FIGS. 33, 34, 35 and 36 are diagrams illustrating examples of a TD method.

FIG. 37 is a diagram illustrating a management table regarding the number of structural elements.

FIG. 38 is a diagram illustrating a target-language sentence pattern having a degree of importance.

FIG. 39 is a diagram illustrating an example of a target-language sentence pattern having structural elements in the target-language.

FIG. 40 is a diagram illustrating a translation thesaurus dictionary.

FIG. 41 is a diagram illustrating a thesaurus dictionary in the target-language.

FIGS. 42 and 43 are diagrams illustrating examples of a target-language phrase pattern database having a difference pattern.

FIG. 44 is a diagram illustrating examples of an original sentence (J).

FIG. 45 is a diagram illustrating examples of translated sentences (E) in difference pattern.

FIG. 46 is a diagram illustrating examples of a translated sentence (E).

FIG. 47 is a diagram illustrating an example of a structure conversion rule of the target-language intermediate structure.

FIGS. 48, 49, 50, 51, and 52 are diagrams illustrating examples of the interaction with a user through a screen.

FIG. 53 is a diagram illustrating a guidance list.

FIG. 54 is a diagram illustrating another example of an original sentence.

FIG. 55 is a diagram illustrating an example of the source-language intermediate structure.

FIG. 56 is a diagram illustrating an example of contents of the request information buffer.

FIG. 57 is a diagram illustrating an example of the guidance list database.

FIG. 58 is a diagram illustrating an example of an interactive screen.

FIG. 59 is a diagram illustrating another example of contents of request information buffer.

FIG. 60 is a diagram illustrating another example of an interactive screen.

FIG. 61 is a diagram illustrating an example of a screen on which an essential part and an optional part are displayed so as to be distinguished from each other.

FIG. 62 is a diagram illustrating examples of an optional phrase pattern.

FIG. 63 is a diagram illustrating an example of a phrase pattern recorded after a confirmation operation.

FIG. 64 is a diagram illustrating examples of the phrase pattern having a classifying field.

FIG. 65 is a block diagram illustrating a language conversion system according to another embodiment of the present invention.

FIG. 66 is a diagram illustrating a structure of a translation unit shown in FIG. 65.

FIG. 67 is a diagram illustrating a structure of a result confirmation unit shown in FIG. 65.

FIG. 68 is a diagram illustrating an example of information flows in the language conversion system.

FIG. 69 is a flow chart illustrating an operation of the language conversion system.

FIG. 70 is a diagram illustrating an example of the extraction of the intermediate structure.

FIG. 71 is a diagram illustrating an example of the IC candidate list database.

FIG. 72 is a diagram illustrating an example of an IC candidate correspondence database.

FIG. 73 is a diagram illustrating an example of the sentence pattern database.

FIG. 74 is a diagram illustrating an example of the request information acquisition.

FIG. 75 is a diagram illustrating a screen in the original input mode.

FIG. 76 is a diagram illustrating a screen in the result confirmation mode (the SD translation).

FIG. 77 is a diagram illustrating a screen in the result confirmation mode (the TD translation).

FIG. 78 is a diagram illustrating a screen in the result confirmation mode (the word-for-word translation).

FIG. 79 is a diagram illustrating a resolution selecting screen in the result confirmation termination mode.

FIG. 80 is a block diagram illustrating a language conversion system according to another embodiment of the present invention.

FIG. 81 is a diagram illustrating an original, morphemes and analyzed structure.

FIG. 82 is a diagram illustrating an example of contents of the IC "test" in the IC dictionary.

FIG. 83 is a diagram illustrating examples of the IC representation pattern in the IC representation pattern dictionary.

FIG. 84 is a diagram illustrating examples of the IC representation pattern in the IC representation pattern dictionary, each pattern having the source-language pattern.

FIG. 85 is a block diagram illustrating a language conversion system according to another embodiment of the present invention.

FIG. 86 is a flow chart illustrating an operation of the language conversion system.

FIG. 87 in parts of (a), (b), (c), (d) and (e) is a diagram illustrating an example of a process for the target-language pattern.

FIG. 88 is diagram illustrating an example of an interaction screen.

FIG. 89 is a flow chart illustrating another operation of the language conversion system.

FIGS. 90, 91 and 92 are diagrams illustrating examples of judging information.

FIG. 93 is a flow chart illustrating another operation of the language conversion system.

FIG. 94 is a flow chart illustrating another operation of the language conversion system.

FIGS. 95, 96, 97, 98 and 99 are examples of the interactive screens.

FIG. 100 is a block diagram illustrating a text creating system according to an embodiment of the present invention.

FIG. 101 is a flow chart illustrating an operation of the text creating system.

FIG. 102 is a diagram illustrating a text in the source-language (an original text).

FIG. 103 in parts of (a), (b) and (c) is a diagram illustrating elements of the original text, target-language element corresponding to the elements and a typical text adaptive to a set of the elements.

FIG. 104 in parts of (a) and (b) is a diagram illustrating structural elements of the typical text and created sentences corresponding to the structural elements.

FIG. 105 is a diagram illustrating a result in which request information extracted from the original text is inserted into the created text.

FIG. 106 is a diagram illustrating a text edited by a user.

FIG. 107 is a diagram illustrating an example of a text structural element database.

FIG. 108 in parts of (a), (b) and (c) is a diagram illustrating examples of the typical text database.

FIG. 109A is a diagram illustrating an example of contents of the text element database.

FIG. 109B is a diagram illustrating an example of contents of the text structure database.

FIG. 109C is a diagram illustrating an example of contents of the sentence pattern database.

FIG. 110 is a flow chart illustrating an operation of the text creating system according to another embodiment of the present invention.

FIGS. 111A and 111B are diagrams illustrating examples of the user-interface when the user selects text elements.

FIGS. 112A, 112B and 112C are diagrams illustrating examples of sets of selected elements.

FIGS. 113A and 113B are flow charts in a process carried out by the system.

FIG. 114A is a diagram illustrating an example of a created text.

FIG. 114B is a diagram illustrating an example of text edited by the user.

FIG. 115 is a block diagram illustrating a text creating system according to another embodiment of the present invention.

FIG. 116 is a diagram illustrating the text element database.

FIG. 117 is a diagram illustrating the text structure database.

FIG. 118 is a diagram illustrating the text structure database.

FIG. 119 is a diagram illustrating the sentence pattern database.

FIG. 120 is a diagram illustrating an original sentence.

FIG. 121 is a diagram illustrating a set of text elements of the original sentence.

FIG. 122A and 122B are diagrams illustrating examples of a created text.

FIG. 123 is a flow chart illustrating an operation of the text creating system according to another embodiment of the present invention.

DESCRIPTION OF THE PREFERRED EMBODIMENTS

A description will not be given of an embodiment of the present invention.

FIG. 2 shows the appearance of a system according to the embodiment of the present invention. Referring to FIG. 2, the system has a display unit 1, a floppy disk unit 2, a keyboard unit 3, a printer 4 and a scanner 5 (including an OCR). A housing of the keyboard unit 3 houses also circuit boards for a system controller, a text editing unit, various data bases, a communication system and other functions. The OCR (Optical Character Reader) included in the scanner 5 optically reads images on a sheet and converts image data into character codes. The character codes are used as input text data. The floppy disk unit 2 stores texts described by character codes, the texts being used as input texts. Output texts may be also stored in the floppy disk unit 2. Another storage unit, such as a magnetic tape unit or an optical disk unit, may be used as a recording medium for texts. The printer 4 prints texts (input texts, output texts, an original, a translation, an intermediate structure of a sentence and the like). The keyboard unit 3 is used to make texts, to input various instructions and data required for the text editing. Instructions for the printer 4 and the scanner 5 are also input by the keyboard 3. The display unit 1 displays texts input by the keyboard unit 3, an original, an intermediate structure of a sentence, a translation, a text to be output, contents of a translation knowledge data base, image data supplied from the scanner 5, information required by the system and other various information items. A housing of the keyboard unit 3 houses various circuit boards on which a controller, a text editing unit, a communication unit and the like are formed.

The system is formed as shown in FIG. 3.

Referring to FIG. 3, the display unit 1, the floppy disk unit 2, the keyboard 3, the printer 4 and the scanner 5 are coupled to a system controller. A text editing unit 6 and a communication unit 13 are also coupled to the system controller. The communication unit 13 carries out communication with external unit via a telephone line or a personal line. In the communication, the communication unit 13 receives character codes used as input text data. The communication unit 13 may transmit character codes of an output text.

The text editor unit 6 edits an original text and a translated text. The editing unit 6 is formed of a translation editor 7, a original text data base 9 and a translated text database 12. The translation editor 7 includes a translation body 8, a translation knowledge database 10 and a dictionary editor 11. The text editing unit 6 extracts a part to be translated from a original text, supplies the extracted part to the translation editing unit 7 and receives a translation from the translation editing unit 7. The translation editing unit 7 manages the translation body 8 for translating a original text. The dictionary editor 11 is positioned at the same level as the translation editor 7 under the text editing unit 6, can input data to the translation knowledge database 10, and can change and edit contents of the translation knowledge database 10.

Original texts and original sentences supplied from the scanner 5 and the floppy disk unit 2 are formatted by the translation editor 7, the formatted original texts and sentences are stored in the original text database 9. Texts and sentences translated by the translation body 8 and texts processed by the translation editor 7 are stored in the translation text database 12. The text editing unit 6 has pointers indicating correspondence between contents (texts, chapters or sentences) in the original text database 9 and the translation text database 12.

In the translation body 8, a data structure is changed as shown in a left side area of a block having the reference number "8", and functions are provided as shown in a right side area of the block. In the block indicating the translation body 8, data items and functions put in brackets are not always required for the present invention. The translation knowledge database 10 is a general term of a set of a analysis grammar/dictionary database, a field classified keyword list database, an IC candidate list database, a target-language pattern database, a translation dictionary database and the like.

FIG. 4 shows relationships between functions in the translation body 8 shown in FIG. 3. In FIG. 4, a keyboard/display represents the translation editor 7, and other parts are in the translation body 8 (an original, a translation and an input IF are located in a boundary between the translation editor 7 and the translation body 8).

Symbols in FIG. 4 will be described now.

Terms sandwiched by .vertline. .vertline. represents processing function. .smallcircle.is a information node in a translation process, and corresponds to a information memory receiving and transmitting information between respective processing functions. () has the same function as .largecircle., and information in each memory indicated in () is frequently rewritten. --.fwdarw. and .rarw..fwdarw. indicate flows of information in the translation body 8. =.fwdarw. indicates a flow of information between the translation body 8 and an external unit. In FIG. 4, to avoid the complexity, lines indicating controls between respective processing function are omitted.

A description will now be given of each of the processing functions in a case where the source-language is Japanese (an original sentence is a Japanese sentence) and the target-language is English (a translated sentence is an English sentence).

ANALYSIS

An original text (J) is analyzed sentence by sentence. Each original sentence are analyzed using analysis grammar roles and analysis dictionary in the analysis grammar/dictionary database. As a result, an intermediate structure (S) in the source-language is obtained. In a mode in which a translation unit will be described later is "batch 1", a dividing process for a sentence is carried out in earlier steps of this analysis process.

FIELD DETERMINATION

A field of a text to be translated is automatically determined, contents of the intermediate structure (S) in the source-language is analyzed using field classified keyword list database. Thereafter, processes are performed in the determined field. This function is not essential in the present invention. The field nay be identified by a user.

IC EXTRACTION

IC is a central term of an idea to be explained by a sentence, and is an abbreviation of an idea center. In a sentence, for example, "we want to discuss an activity theme hereafter", although a main verb is "want", the subject matter of the sentence is "discuss". In addition, in a sentence "it is regarded that this contract is approved when our firm receives the application money", although the main verb is "regarded", the subject matter of the sentence is "approved". That is, "discuss" and "approved" are respectively IS (idea centers) of the above sentences. In this embodiment, to make decision the IC easy, the IC candidate list database is used. Terms to be the IC are extracted from the intermediate structure (S) of an object sentence with reference to the contents of the IC candidate list database, and the extracted terms are stored in an extracted IC buffer. Each of the extracted IS has a pointer for the target-language sentence pattern which is written in the IC candidate list database. This function (IC EXTRACTION) is not essential in the present invention. The IC can be decided by multiple use of the following function (TARGET-LANGUAGE PATTERN EXTRACTION).

TARGET-LANGUAGE PATTERN EXTRACTION

The contents of the target-language pattern database are retrieved by using of the pointer in the extraction IC buffer, the intermediate structure (S) is analyzed in accordance with retrieved contents, and the IC of the object sentence is temporarily decided. A word (or idiom), in the target-language (English in this case), corresponding to the IC which has been temporarily decided is selected. The structure of a sentence is decided by phrase patterns identified by the target-language sentence pattern which is required by the above selected word corresponding to the IC. On the other words, it is analyzed what part the English IC fulfills in the target-language sentence and what words, phrases and clauses the IC requires, the request information is extracted from the intermediate structure (S), and the structure of the sentence in the target-language is decided. In this case, the structure of the sentence is according to the target-language, and information in the source-language is made so as to correspond to information in the target-language. The information obtained by this function is stored in a request information buffer. In the system which is not an interactive conversation type, a default value is substituted for information which has not been extracted from the intermediate structure (S), and a process corresponding to this function is completed. In the system in which translation body 8 is operated with interactive conversation with an operator, if there is no information needed to decide the structure of the sentence in the information extracted from the intermediate structure, information to be requested of the operator is prepared. Comment information indicating what quality the information to be requested of the operator has is also prepared. Furthermore, comments indicating the reason why this IC has been selected or the meaning of selection of this IC are supplied to a guidance buffer.

GUIDANCE INFORMATION EXTRACTION

Regarding difference information indicating the difference between the original having no comment in the guidance buffer and translation, comments corresponding to the difference information is read out from the guidance list database, and the comments fills up the lack of the information in the guidance buffer.

In this embodiment, with respect to the original, the reason why the sentence pattern is used, and the comments regarding essential words are described in the target-language sentence pattern. Theses information items may be recorded in the guidance list and be read out therefrom.

DISPLAY SHAPING

The original (J), the contents of the request information buffer (including the translation (E), reference sentences) and the contents of the guidance buffer are composed and displayed on the display unit 1. The information to be displayed is formed so that the user can easily recognize it (the system is provided with a pointer management function by which the corresponding relationship between information input by the keyboard 3 and the request source of the information are maintained). In addition, the information to be displayed is changed in accordance with a rewriting request, and the information in the display buffer is rewritten.

INPUT INTERFACE (IF)

When the operator should correct the information in the display buffer looking the screen of the display unit 1, and when the operator inputs the request information in accordance with the request from the system, the information items are input from the keyboard 3. At this time, the interface between the input information and the display formation are carried out.

TRANSLATION STATE MEMORY

The operator identifies a state under which the translation body processes a sentence to be translated, the identified state is stored in a flag memory. If a field is automatically set in the system, the determination result of the field is stored in the system. The translation process may be carried out in accordance with the field stored in the system.

SHAPING 1

Only information needed for the translation sentence (E) is read out from the request buffer, and an target-language intermediate structure (T) capable of easily creating the translation sentence is generated. In the interactive mode, the pointer information is also shaped to maintain the relationship between the processed information and the information supplied to the "DISPLAY SHAPING".

SHAPING 2

Only information needed for the reference sentence (J') (a sentence written by the same language of the original sentence) is read out from the request buffer, and a reference intermediate structure (S') capable of easily creating the reference sentence is generated. In the interactive mode, the pointer information is also shaped to maintain the relationship between the processed information and the information supplied to the "DISPLAY SHAPING".

CREATION 1

The translation sentence (E) in the surface level of the target-language is created from the target-language intermediate structure. In the interactive mode, the pointer information is also created to maintain the relationship between the created information and the information supplied to the "DISPLAY SHAPING".

CREATION 2

The reference sentence (J') in the surface level of the source-language (the Japanese) is created from the reference intermediate structure. In the interactive mode, the pointer information also created to maintain the relationship between the created information and the information supplied to the "DISPLAY SHAPING".

T-SHAPING

The relation pointer with respect to the information in the "DISPLAY SHAPING" is removed.

A description will now be given of databases used in the various functions (processes).

ANALYSIS GRAMMAR/DICTIONARY DATABASE

Since contents of this database are not directly concerned with the present invention, the detailed description is not given. This database may be selected from among conventional databases, but the following condition must be satisfied. The dictionary is constituted so that a writing form of each word in the source-language intermediate structure (S) is normalized, and in the analysis process, each word having the normalized notation is processed. For example, "" (okonau) can be written as "" (okonau) and "" (okonau). Even if any of the three words is used in the original sentence, the word is uniformly written as "" (okonau) in the source-language intermediate structure (S). Regarding sets of words {"" (todomeru), "" (todomeru), "" (todomeru)}, (todomeru), {"" (sokubunsuru), "" (sokubunsuru), "" (solubunsuru)}and {"" (shinpojumu), "" (shinpogiumu)}, even if any of the words in each set is used in the original sentence, one of the words in each set is representatively used in the system. Hereinafter, underlined characters as indicated above correspond to kanji.

Other databases will be described below with reference to FIGS. 5-14.

FIELD CLASSIFIED KEYWORD LIST DATABASE

FIG. 5 shows an example of the field classified keyword list database.

In FIG. 5, words by which a field can be easily decided are gathered, and these words are mainly noun. To characterize the field, words are gathered without distinction between the short unit and the long unit. This database is provided with, as the attribute, "ENTRY", "READING", "PART OF SPEECH", "FEATURES", and "USING FREQUENCY" in each field. The words in the "ENTRY" are written in the representative form in the same manner as those in the intermediate structure (S). The "READING" and the "PART OF SPEECH" are used, in a case where this database is retrieved by the dictionary editor, to correct information in this database and to add information to this database. Even if words have the same notation, their meanings may differ from each other in accordance with fields in which the words are used. Thus, the "FEATURE" is used to distinguish meanings of the words having the same notation. For example, a case where a word "ukekuchi is used as a mouth is identified by HEL in the "FEATURE", and a case where a word "ukekuchiis used as an inlet of a mailbox is identified by IEL in the "FEATURE". Using frequencies of words are statistically investigated in respective field, and the values obtained by the investigation is logarithmically normalized. The "USING FREQUENCY" has the normalized values concerning the using frequency of words. This database is provided with three fields, the first field that words have been investigated in newspapers and magazines, the second field that words have been investigated in papers regarding software and hardware of computers, and the third field that words are investigated in trade journals regarding the heavy electrical equipment. The number of fields is not limited and the "USING FREQUENCY" may be roughly or finely classified into fields.

IC CANDIDATE LIST

FIG. 6 shows an example of the IC candidate list.

A word to be the IC is a declinable word (a verb, an adjective, an adjective verb and the like) and a part of an auxiliary verb ("-da" used for conclusion, "-desu" and "-no youda"). However, a verb "suru" can not be the IC.

This database is provided with, as the attribute, "ENTRY", "READING", "PART OF SPEECH CONJUGATION", "POSSIBILITY" and "POINTER". Each word in the "ENTRY" is written in the present notation used in the intermediate structure (S). The "READING" and the "PART OF SPEECH CONJUGATION" are used, in a case where the dictionary editor retrieves this database, to add information to this database and to correct information in this database. The "PART OF SPEECH CONJUGATION" is also used as an operator of the possibility. The "POSSIBILITY" is a possibility by which a word may be the IC in a Japanese sentence, and is obtained by the analysis of a large number of sentences.

In the analysis of the sentence "kongo no katsudo tema ni tsuite disukasshyon shitai to omoimasu" (we want to discuss an activity theme hereafter.) Words "katsudou suru" (be active), "disukasshyon suru" (discuss) and "omou" (want) can be the IC. The word "disukasshyon suru" (discuss) is the IC in this sentence, so that this word is provided with 3 points. Although the word "omou" (want) is a main verb in this sentence, the "monou" (want) is not the IC in this sentence. Thus, the word "omou" (want) is provided with 0 point. It is not analyzed whether or not the word "katsudo suru" (be active) is the IC in this sentence, so that this word is provided with 1 point.

In another sentence "hon keiyaku wa tosha ga moshikomikin wo juryo shita toki ni seiritu surumonoto shimasu" (It is regarded that this contract is approved when our firm receives the application money), although "shimasu" (regarded) is a main verb in this sentence, the subject matter is "seiritu suru", so that the words "keiyaku suru" (contract), "moshikomu" (apply), "juryo suru" (receive), "seiritu suru" (approve) and "suru" (regarded) are respectively provided with 1 point, 1 point, 1 point, 3 points, and 0 point.

These points are provided to the words with reference to the translated sentence. In a case where the translated sentence in English is a complex sentence, a candidate of the IC corresponding to a word which is the subject matter in the translated sentence in English is provided with 2 points.

The points are totaled every candidates of the IC, and the results are logarithmically normalized. The "POINTER" is used to detect a part corresponding to the IC from the information in the target-language pattern database.

The target-language pattern database is divided into the target-language sentence pattern database and the target-language phrase pattern database.

TARGET-LANGUAGE SENTENCE PATTERN DATABASE

FIGS. 7-11 show examples of the target-language sentence patterns.

In this database, a fundamental sentence pattern for each IC is described in the target-language (English), and a sentence pattern by which contents of the original sentence can be explained easiest is selected as the fundamental sentence. This database has a pointer used to extract phrase patterns, such as a to-infinitive phrase, a gerund phrase, and a preposition phrase.

Referring to FIGS. 7-11, in the first line, the IC, an English term corresponding the IC and the meaning that the IC is used in this case are provided (the English term corresponding to the IC is referred to as an English IC.

<Required>: This label indicates essential information which is information to be essential in the translated sentence having the English IC.

<Preferred>: This label indicates preferred information which is information preferable to be included in the translated sentence having the English IC.

<Optional>: This label indicates optional information which is information apt to be used in the translated sentence having the English IC. Even if the optional information does not exist in the sentence having the English IC, the sentence is not unnatural.

<Subj-.sub.-- >: This label indicates the subject in the sentence.

<DirObj-.sub.-- >: This label indicates the direct object in the sentence.

<IndObj-.sub.-- >: This label indicates the Indirect object in the sentence.

<Xobj-.sub.-- >: This label indicates another object (an object of a to-infinitive phase, an object of a gerund and the like).

<Alternates>: Information to be indicated, as information required in the translation, to the operator and the default information, such as <someone> and <something> are written in the <Alternates>. The information required in the translation is indicated when the information does not exist in the intermediate structure (S), and if the required information is not supplied to the system by the user, the default information is used.

<Feature>: This indicates features of information required. If the features correspond to the information in the intermediate structure (S), the reliability increases.

<Japanese>: This indicates case labels in the intermediate structure (S) and another condition information written in this position.

<English-Pattern-Pointer>: This is a set of pointers indicating phase patterns having contents (arrangement of words or phrases) more detailed than description roughly characterizing the English sentence pattern (values of <Restriction>). This is indicated, for example, as <v-.vertline.test.vertline.-001>.

TARGET-LANGUAGE PHRASE PATTERN

FIG. 12 shows an example of the target-language phrase pattern.

A sentence structure is decided by use of phase patterns identified by the target-language sentence pattern so that whole contents explained by the source-language can be explained by the target-language.

<Pattern>: This is a unit of the arrangement of English words, default values and other phrases which are grammatically acceptable in the sentence.

<Example>: This is an example of explanation used for the update of this database.

<Component>: This is indicated by <Xn, Xm>. <Xn, Xm> means that the m-th X identified by the n-th X is selected, X being an element described in the <Pattern>.

TRANSLATION DICTIONARY

FIG. 13 shows an example of the translation dictionary.

Independent words or content words in the source-language are provided with equivalents in the target-language. Even if the equivalents regarding the same words, they may differ from each other in accordance with the fields. The translation dictionary may have the function of the field classified keyword list described above, but in the light of the simplicity of the management of the databases, it is preferable that the translation dictionary be separated from the field classified keyword list.

The translation dictionary is provided, as the attribute, with "ENTRY", "READING", "PART OF SPEECH", "FEATURES", and "ENGLISH DESCRIPTION". In the "ENGLISH DESCRIPTION", English equivalents of Japanese words are described. The "FEATURES" is used for the relationship analysis with respect to features required in the target-language phase patterns. Other items are the same as those in the field classified keyword list.

GUIDANCE LIST

FIG. 14 shows an example of the guidance list database.

This list has comments why the sentence pattern is selected for the original, why equivalents corresponding to words in the original do not exist in the translated sentence, and why equivalents corresponding to words which do not exist in the original is required.

EXCHANGE: This is a comment given in a case where appropriate terms are substituted for terms which are generally easily erroneous for Japanese.

REQUIRED: This is a comment given in a case where words corresponding to words not described in a Japanese sentence are needed in English.

RELATIONAL: This is a comment given in a case where an English word to which a word (especially an independent word) described in a Japanese sentence is to be directly translated does not exist in a translated sentence.

An operation of the translating system according to the present invention will be described below with reference to FIGS. 15-26.

First, the operator must decide a state in which the translation body 8 is to be used. This system must have a memory for storing the state of the translation body. FIG. 15 shows a translation state memory.

Information indicating the state in which the translation body 8 is to be used and information required when the translation body 8 changes conditions are stored in the translation state memory. Flags having "0" or "1" are assigned to state items in the translation state memory. The state items are "TRANSLATION UNIT", "INTERACTION" and "FIELDS". The state item "TRANSLATION UNIT" is divided into "BATCH 1", "BATCH 2" and "SINGLE SENTENCE UNIT". The "BATCH 1" is a mode in which an automatic sentence separation process for separating sentences in a text from each other can be performed within the analysis process so that a text having a plurality of sentences can be translated. The "BATCH 2" is a mode in which a set of sentences clearly separated from each other is translated. "SINGLE SENTENCE UNIT" is a mode in which a single sentence is translated. The state item "INTERACTION" is divided into "EXISTENCE OF INTERACTION", "EXISTENCE OF GUIDE DISPLAY", "EXISTENCE OF WORD LEARNING" and "EXISTENCE OF SENTENCE PATTERN LEARNING". The state item "FIELD" is divided into four fields (0)-(3) in this embodiment. The number of fields is not limited to four. In this embodiment, the field (0) corresponds an automatic mode, the field (1) is a generic field, the field (2) is related to the electronics and information, and the field (3) is related to the heavy electrical equipment.

A manual flag section is a flag section in which flags can be identified by the operator. An automatic flag section is assigned to only the state item "FIELD". The manual flag section is prior to the automatic flag section, and the automatic flag section is activated in only a case where a flag of a field (0) in the manual flag section is set at "1". In this embodiment, it is assumed that the state items "SINGLE SENTENCE UNIT", "EXISTENCE OF INTERACTION", "EXISTENCE OF GUIDE DISPLAY", and the field (0) are active (corresponding flags in the manual flag section are and the items "EXISTENCE OF WORD LEARNING" and "EXISTENCE OF SENTENCE PATTERN LEARNING" are inactive (corresponding flags in the manual flag section are "0").

Hereinafter, a description will be given of operations for translating two original sentences (1) and (2) shown in FIG. 16. In the "SINGLE SENTENCE UNIT", the sentences are read out from the original text database sentence by sentence. Each of the sentences read out from the original text database is assigned with a tag of SourceTex (J).

FIG. 18 shows an example of the source-language intermediate structures (S) obtained by the analysis of the original sentence (J).

Referring to FIG. 18, in an item "sem", an independent word (including an affix and a coined word component) which is the main analysis result is described in the original form. A state in which the independent word is used in the sentence is indicated immediately under the item "sem". The state may be indicated using a plurality of lines if necessary, and the state may be not indicated if unnecessary. An item name positioned at the front of › ! is a symbol indicating a part, in the sentence, assigned to contents in the › !, and is referred to as a part label.

Here, it is indicated that a word "okonau" is the main verb of the sentence (1). "tens" and "PAST" (ta) indicate that the word is accompanied with an auxiliary verb "ta" so as to be in the past tense. Although a word "testosuru" is originally a sa-hen verb, it is indicated that the word "testosuru" is used as a noun in the sentence. "wo obj" positioned at the front of › ! in which "testosuru" is written indicates that the word "testosuru" is used as the object in the sentence. A part label "pre-supp" is assigned to a set of words "suru", "taisho" and "hendensho".

In the intermediate structure (s) of the sentence (2), since the main verb "suru" is the present tense, the "tense" and "PASTE" are omitted as the default. "aspect" and "STATE" (iru) indicates that the sentence is a state indicating sentence accompanied with an assistance verb "iru". <renyo>mod indicates that a set of words "doyoda, "baai", "shorisuru" and "togo" in the › ! modifies a word in a renyo-style (hi) of the adjective verb ". . . da". The "ha topic" is the part label indicating that a set of words "jitsugensuru", "jo", "konpyuta" is the subject matter of the sentence.

When the analysis process is completed, it is determined, with reference to the state item "FIELDS", whether or not the automatic mode is active. If the automatic mode is active (the flag of the field (1) is set to "1"), the field determination process is carried out. If a field has been designated, the field determination process is skipped.

In a case of the sentence (1), the values "okonau", "tesutosuru", "suru", "taisho", "hendensho" in all the items "sem" are extracted from the intermediate structure (S). The "ENTRY" of the field classified keyword list (see FIG. 5) is searched for the above values. Only the value "hendensho is detected, and a frequency "076" in the field (1), a frequency "014" in the field (2) and a frequency "232" in the field (3) are read out from the "FREQUENCY". The frequency in the field (3) has the largest value, and this information is supplied to the translation state memory, so that the automatic flag corresponding to the field (3) in the state item "FIELDS" in the translation state memory (see FIG. 15) is set to "1". Hereinafter, information belonging to the field (3) related to the "heavy electrical equipment" is preferentially processed.

In a case of the sentence (2), all the values in the item "sem" are extracted from the intermediate structure (S), and the classified keyword list database is searched for the extracted values. As a result, values "gengo" having frequencies "045", "224" and "006" in the fields (1),(2) and (3), "ronrigata" having frequencies "021", "088" and "014", "besu" having frequencies "156", "045" and "092", "baai" having frequencies "273", "210" and "084", "togo" having frequencies "000", "043" and "000" and "konpyuta" having frequencies "033", "321" and "174" are detected in the "ENTRY" of the classified keyword list database. The total values of the frequencies of the extracted values in the fields (1), (2) and (3) are then calculated. The total values of the frequencies are "528" in the field (1), "931" in the field (2) and "370" in the field (3)>. The total value of the frequencies in the field (2) is maximum. Thus, the flag corresponding to the field (2) in the automatic flag section of the translation state memory is set to "1". Hereinafter, information belonging to the field (2) related to the "electronics/information" is preferentially processed.

When the field determination process is completed, the IC is extracted from a sentence to be translated.

In a case of the sentence (1), values (words) "okonau", "testosuru", "suru", "taisho", "hendensho" in all the items "sem" are extracted from the intermediate structure (S) (the part label belonging to each item "sem" is also extracted), and the "ENTRY" of the IC candidate list (see FIG. 6) is searched for the extracted values. In this case, "testosuru" (wo obj) and "okonau" (Hed) are detected. The possibility E=f(S,P) by which each detected word is the IC is calculated using a coefficient P depending on the inflection of the part of speech of each detected word and the part label and a possibility value P. The calculation results are stored as scores in the extraction IC buffer along with pointers, as shown in FIG. 17.

In a case of the sentence (2), all the values (words) in the items "sem" and the part labels are extracted from the intermediate structure (S), and the IC candidate list (see FIG. 6) are searched for the extracted values in the items "sem". In this case, two words "kotsigemsuru" (wa topic) and "shorisuru" (<renyo>mod) are detected, and the possibility E by which each word is the IC is calculated. The calculation results are stored as scores in the extraction IC buffer along with pointers, as shown in FIG. 17.

After this, the target-language pattern of each sentence is extracted.

In a case of the sentence (1), the target-language sentence pattern database is searched for the pointer "0378" of the word "testosuru" having the highest score in the extraction IC buffer, and information shown in FIGS. 7-11 is obtained.

First, the intermediate structure (S) is checked for the description of the <Restrictions> in the <English-Pattern-Pointer>. In this case, it is checked where the part label of the DirObj-PATIENT exists in the intermediate structure (S). However, since the DirObj-PATIENT does not exist in the intermediate structure (S), the process proceeds to the next step. In item "AND", (@noun, sem (suru)--upper structure, function name (wo obj)) is described. This means that the word "tesutosuru" has the attribute "noun", the function name thereof is "wo obj" and "suru" exists in the upper item "sem". As a result of checking, conditions of the "@noun" and the "function name" (wo obj) are satisfied, but the "suru" does not exist in the upper item "sem". Next, it is checked whether or not "ukeru" exists in the upper item "sem". However, the "ukeru" does not exist in the upper item "sem".

In the next item "AND", (@noun, sem (okonau),--upper structure function name (wo obj)) is described. The intermediate structure (S) is checked in the same manner as that in the above case. As a result, it is determined that the word "testosuru" has the attribute "@noun", the "okonau" exists in the upper item "sem" and the part label is (wo obj). The all the conditions are satisfied, so that this pattern is accepted. Since the pointer is <v-.vertline.test.vertline.-003>, conduct <n-.vertline.test.vertline.-002> is obtained due to the retrieving of the target-language phrase pattern database. This means that the translated sentence should have a sentence pattern in which a word "conduct" having a direct object "test" is a main verb. Information in < > is a pointer, and the retrieval process is performed using the information in < >. As a result, .vertline.test.vertline.(<p-.vertline.of.vertline.-001>) and (<p-.vertline.for.vertline.-001>) are obtained. This description has a component section in which (X1, X1)<DirObj-PATIENT> and (X2, X2)<Xobj-CONDITION> are indicated. Thus, <p-.vertline.of.vertline.-001> is retrieved, and .vertline.of.vertline.<SOMETHING> is obtained as the default value.

After this, the intermediate structure (S) is searched for a word using the part label of <DirObjPATIENT>. As this label is an English label, the English label is converted into a Japanese label using a reference table in which one or a plurality of Japanese labels corresponding to the English label are arranged in the priority order. As a result, "wo obj" is obtained the highest priority. The intermediate structure (S) is retrieved using this label, a word "hendensho is obtained. Here, the translation dictionary database is searched for hendensho, so that an equivalent "transformer station" in English is obtained, the attribute of the equivalent being an INS (institution). As the INS belongs to the PATIENT, it is confirmed that the "transformer station" may be inserted after .vertline.of.vertline..

When <p-.vertline.for.vertline.-001> is checked, .vertline.for.vertline.<SOME PURPOSE> is obtained as the default information. A Japanese label corresponding to the part label of <Xobj-CONDITION> is "ni obj", and the intermediate structure (S) is searched using the label "ni obj". As a result, a word "taisho" is obtained. Here, the translation dictionary database is retrieved, so that an equivalent "subject" and a meaning attribute "APP()" are obtained. Further, as the CONDITION is not directly related to the APP, it is determined that the word taisho is not inserted after the .vertline.for.vertline.. Thus, the taisho is not selected as a candidate of <SOME PURPOSE>. After the information extracted from the intermediate structure (S) in accordance with the request from the target-language pattern is completely processed, the item "sem" (suru) is processed in a "specific process of suru".

To be concrete, if the item "sem" is left over in a pattern ". . . a postpositional word (functioning as an auxiliary)/"APP":meaning appanage/suru . . . ", such as ". . . wo/taisho/nishite . . . ", ". . . wo/taisho/toshita", ". . . no/baai/nishite . . . " and ". . . wo/baai/toshite", the word "suru" is not converted into a word in the target-language. In a case where a Japanese word is not directly converted into an English word as described above, the information regarding this is supplied to the guidance information extraction. At the last, returning to the target-language side, the contents of <Required> are checked. As a result, it is determined that the item <Subj-AGT> is not filled with information, and the contents (the default value of <someone>) is supplied to the request information buffer. The comments are transmitted to the guidance buffer. The item <DirObj-PATIENT> is provided with a word "test" itself and "PreDO" of the word "test".

Now, the whole information in the intermediate structure of the sentence (1) has been completely processed. However, even if the process for the information in the intermediate structure (S) reaches an impasse, the target-language pattern extraction process may be terminated at a time the patterns in English have been completely used (reference to FIGS. 19 and 20).

The guidance information extraction receives from the target-language pattern extraction information indicating that the items "sem" (taisho) and (suru) are not necessary. Thus, "taishonishite" is created from the structure, and a guidance message indicating that the meaning is included in the sentence pattern is selected from the guidance database. The "taishonishite" and the guidance message are supplied to the guidance buffer.

In the display shaping process, <someone> and <who> in the item <Subj-AGT> are selected as the essential information from the request information buffer. The label indicating these are essential information and the essential information are supplied to the display buffer. To be concrete, <<someone: >> as the information for IM(e) and <<who: >> as the information for IM(J') are respectively supplied to the IM(e) and IM(J').

In addition, comment information--"tesutosuru": "test": "in a case where machines and equipments are tested"--is selected from the guidance buffer, and are shaped into a form so as to be easily read by the operator. The comment information is supplied to the Comment section of the display buffer. Further, the comment information of "taishonishite" which has been supplied from the guidance information extraction is selected from the guidance buffer, and is supplied to the Comment section of the display buffer. In addition, the information written in the <Alternates> which is the essential information in the target-language pattern extraction is selected from the guidance buffer, and is supplied to the Example section of the display buffer.

DISPLAY BUFFER

In FIG. 22, the description surrounded by › ! in the item value means that information identified by the title is indicated. The description without › ! is the description itself. The description in the item "Intoed(e):" is surrounded by › !, contents of the description is "GenResult(E')". Thus, a translated sentence (E') at this time is displayed. This is a sentence created by shaping of the information read out from the English section of the request information buffer. This has a pointer for indicating the relationship between the information in the request information buffer and the translated sentence in the surface level. The description in the item "Intoed(J'):" is surrounded by › !, contents of the description is "GenResult(J')". Thus, a reference sentence (J') at this time is displayed. This is a sentence created by shaping of the information read out from the Japanese section of the request information buffer. This has a pointer for indicating the relationship between the information in the request information buffer and the reference sentence in the surface level.

Text(J) indicates that SourceTex(J) is read out from the original memory and is supplied to the translation editor 7.

Text(e) indicates that TranTex(E) is read out from the translation memory and is supplied to the translation editor 7.

SHAPING 1

In the request information buffer, the intermediate structure described by English sentence patterns are provided along with corresponding information of a Japanese sentence. only information items needed to create an English sentence is selected from the information items in the request information buffer, and are shaped so as to easily create the English sentence. The headings in the request information buffer are described in Japanese and in English. "conduct a test" in English is selected from the headings in the request information buffer, the main verb "conduct" is a value of the Head-label, and the direct object "a test" is a value of the DirObj-label. In this case, these are associated with the contents of the PrepDo-label in the request information buffer. The value of the Xobj-label may be moved to the creation side as it is. In these operations, the information in Japanese is removed. The results are indicated as shown in FIG. 22 (1).

A sentence in the surface level is created in accordance with creation rules in the "CREATION 1".

The "tense" label is "PAST", and the main verb is changed to the past tense. The essential value of the subject word is surrounded by << >>. If the essential value does not exist, the value (a word) of the DEF-label is inserted into << >>. If the essential value exists, the value of the DEF-label is first inserted into << >> and a value (a word) of the sem-label is then surrounded by (). A value (a word) of a phrase such as the Xoj-label is surrounded by < >. At this time, "<<someone>> conducted a test of <<something (the transformer station)>> <for some purpose>" is obtained. This sentence label is the GenResult (E').

In the T-shaping process, accompanying information of a part having no vagueness is removed from the sentence labeled by the GenResult (E'), and a sentence having a form close to the normal description form (If there is no vagueness at all, a sentence having the normal description form is obtained). At this time, "<<someone>> conducted of a set of transfer station".

SHAPING 2

The request information buffer is provided with the intermediate structure described by English sentence patterns. The request information buffer is also provided with corresponding information concerning the Japanese sentence. Only information needed to create a Japanese sentence is read out from the request information buffer and is shaped in a form by which the Japanese sentence is easily created. The "SHAPING 1" can be actualized by a simple shaping process, but a creating process is required for the "SHAPING 2" in a some measure. The headings in the request information buffer are described in English and in Japanese, "testo wo okonau" in Japanese is selected from the headings in the request information buffer. The main verb "okonau" is a value of the sem-label of Head, and the direct object "testo" is a value of the obj-label. In this case, these are associated with the contents of the PrepDo-label in the request information buffer in the same manner as those in the above English case. Now, a value of the candidate-label is "hendensho and a value of the Prep-label is "of", so that the "no"mod-label is added to the structure. A value of the Xobj-label may be maintained as it is, but the form thereof is converted into that in Japanese. Since a value (a word) of the option label is "nannotameni" and a value of the Prep-label is "for", the mod-label is added to the structure. The process is completed at this time, and the result including the option label may be moved to the creation side. All the English information is removed from the result, and the result is described as shown in FIG. 24 (1).

A sentence in the surface level is created in accordance with creation rules in the "CREATION 2".

The "tense" label is "PAST", and a past auxiliary verb "ta" is substituted for an auxiliary verb.

The essential value, for the subject, is described so as to be surrounded by << >>. If there is no value to be in << >>, a value of the essential label is entered in << >>. If there is a value to be in << >>, the candidate label is substituted for the label name and the value is entered in << >>. An optional clause such as the mod-label is described such that a value of the option label is surrounded by << >>. At this time, "<<darega>><nannotameni><<hendensho>>no testo wo okonatta" is obtained. The label of this sentence is the GenResult (J') (see FIG. 26 (1)).

A description will now be given, with reference to FIGS. 27-30, of screens on the display unit.

When the initial setting operation (see FIG. 15) is completed and an original sentence is selected, the display unit has a screen as shown in FIG. 27. In this state, when a starting instruction of the translation is input from the keyboard, processes from .vertline.ANALYSIS.vertline. to .vertline.CREATION 2.vertline. and .vertline.T-SHAPING.vertline. are performed. After the processes are completed, the contents in the display buffer is then displayed on the display unit as shown in FIG. 28. Referring to FIG. 28, in a section of the ›editing! reference sentence, a sentence existing at this time in the GenResult (E') is indicated in the "translation (E):" area, and a sentence existing at this time in the GenResult (J') is indicated in the "equivalent (J):" area. In the "comment:" area, a reason why "conduct a test" has been used and a reason why an English word corresponding to a word "taishonishite" in the original sentence does not exist in the translation sentence.

This screen is provided with a message "Please fill the following essential items". An "essential item (Japanese):" and an "essential item (English):" may be filled with words. In "English example:", examples of English words capable of filling the "essential item (English):". When the return key is operated under the condition in which the "essential item (English):" is filled with an English word, the translation sentence (E) and the reference sentence (J') in which the English word in the "essential item (English):" is reflected are created. Thus, contents in the "›English! translation" section and in the "translation (English):" area may be changed, but contents in the "equipment (Japanese):" are not changed. When the "essential item (Japanese):" is filled with a required word, the translation dictionary is retrieved and an English description corresponding to the required word is selected. The English description is used in the following processes. The result as shown in FIG. 29 obtained so that the translation is completed.

After the request information has been obtained, the following operation is carried out.

When a Japanese word "seifu enters the "who" in the Japanese side of the essential subj-label in the request information buffer, the .vertline.TARGET-LANGUAGE PATTERN EXTRACTION.vertline. influences the .vertline.EQUIPMENT CONVERSION.vertline. and the translation dictionary is retrieved. As a result, an equipment "government" of the Japanese word seifu is obtained. The word "government" is inserted into the slot in the English side. The result is shown in FIG. 19 (1)'. After this, in .vertline.SHAPING 1.vertline., .vertline.CREATION 1.vertline., .vertline.T-SHAPING.vertline. .vertline.SHAPING 2.vertline. and .vertline.CREATION 2.vertline., processes are carried out in the same manner as the above case. The results are obtained as shown in FIGS. 25(1)', and 26(1)'.

If the operator who does not satisfy the translation inputs a word "agency" after the word "government" and operates the return key, the contents in ( ) of the request information buffer are changed as shown in FIG. 20 (1)". The processes are then carried out in .vertline.SHAPING 1.vertline., .vertline.CREATION 1.vertline., .vertline.T-SHAPING.vertline.. The respective results are obtained as shown in FIG. 25 (1)". In this case, the contents in the Japanese side of the request information buffer are not changed, so that the reference sentence (J') is not changed. A sentence (1)" in FIG. 26 is equal to a sentence (1)'.

FIGS. 31 and 32 respectively show a second embodiment and a third embodiment of the present invention.

In these embodiment, a translated sentence can be obtained by the same processes as the above embodiment. In addition, FIG. 33 shows an embodiment of a TD method (the source-language, and the target-language pattern database), FIG. 34 shows an embodiment of the TD method (the context), FIG. 35 shows an embodiment of the TD method (with a learning process), and FIG. 36 shows an embodiment of the TD method (with a learning process and the creation of the original text).

A description will now be given of modifications of the pattern matching process regarding the target-language pattern.

It is determined, in accordance with the following methods, how certain the target-language sentence pattern as shown in FIGS. 7-10 is matched to the source-language sentence.

(1) It is determined in accordance with counting matched elements.

The certainty is determined based on the number of elements in the target-language sentence pattern, which elements are satisfied by the source-language information. Examples will be described below.

As to a sentence "hendensho wo taishoni shite, tesuto wo okonatta" (a sentence (1) shown in FIG. 16) (an example a), <Restrictions> of a pattern in the <English-Pattern-Pointers> the IC "testosuru" are compared with the source-language intermediate structure (shown in FIG. 18), and it is determined which <Restrictions> are matched to elements in the source-language intermediate structure. In the embodiment of the present invention, it is determined that (@noun sem (okonau)--upper structure function name (wo obj) is matched to the source-language intermediate structure (FIG. 18) by the following steps 1-8.

In step 1, from among <Restrictions> of the patterns in the <English-Pattern-Pointers>, objects having the sem () item matched to the sem item in the source-language intermediate structure is selected. In the embodiment of the present invention, (@noun sem (okonau)--upper structure function name (wo obj) is matched to the element in the source-language intermediate structure. Furthermore, (wo obj) is also matched to an element in the source-language intermediate structure. Thus, the target-language pattern <v-.vertline.test.vertline.-003> corresponds to the source-language intermediate structure.

In step 2, starting from the target-language pattern <v-.vertline.test.vertline.-003>, target-language phrase patterns <v-.vertline.test.vertline.-003>, <n-.vertline.test.vertline.-002> and <p-.vertline.of.vertline.-001><p-.vertline.for.vertline.001> are successively compared with the source-language intermediate structure in this order.

First, the target-language phrase pattern <v-.vertline.test.vertline.-003> is formed of

    ______________________________________
    (<Pattern>   (conduct <n-.vertline.test.vertline.-002>)
                 (conduct <n-.vertline.test.vertline.-003>)),
    ______________________________________

    ______________________________________
    pattern <n-.vertline.test.vertline.-002> is formed of (<Pattern> (.vertlin
    e.test.vertline. (<p-
    .vertline.of.vertline.-001>) (<vinf-.vertline.to determinel-002>))
    (<Example> "A TEST OF (test-subj) TO DETERMINE (A or B)")
    (<Component>  ((X1 X1)  <DirObj-PATIENT>)
             ((X2 X1)
           <Xnnn-BINARY.sub.-- CONTRAST>)),
    ______________________________________

    ______________________________________
    (<Pattern>   (conduct <n-.vertline.test.vertline.-002>)
                 (conduct <n-.vertline.test.vertline.-003>)
    ______________________________________

    ______________________________________
    (<Pattern>   (conduct <n-.vertline.test.vertline.-002>)
                 (conduct <n-.vertline.test.vertline.-003>))
    ______________________________________

    ______________________________________
    (<Pattern>
              (.vertline.test.vertline.(<p-.vertline.of.vertline.-001>)(vinf-.
              vertline.to determine.vertline.-
    002>))
    (<Example> "A TEST OF (test-subj) TO DETERMINE (A or B)")
    (<Component>
              ((X1 X1)  <DirObj-PATIENT>)
               ((X2 X1)
    <Xnnn-BINARY.sub.-- CONTRAST>)).
    <n-.vertline.test.vertline.-003> corresponds to DirObj-PATIENT -
    "hendensho" : <Required> and Xnnn-BOOLEAN - "anzenka,
    anzendenaika" <Preferred>
    in a case of (<Pattern> (.vertline.test.vertline. (<p-.vertline.of.vertlin
    e.-001>) (vinf-.vertline.to
    determine.vertline.-001>))
    (<Example> "A TEST OF (test-subj) TO DETERMINE (WHETHER A or
    not A)")
    (<Component>
              ((X1 X1)   <DirObj-PATIENT>)
               ((X2 X1)
    <Xnnn-BOOLEAN>)).
    Thus,
    (conduct <n-.vertline.test.vertline.-002>) and
    (conduct <n-.vertline.test.vertline.-003>)
    ______________________________________

    ______________________________________
    (<Required>
             (<Subj-AGT
               20
             (<DirObj-PATIENT>
             30
    (<Preferred>
             (<Xobj-CONDITION>
             3
             (<X-BINARY-CONTRAST>
             (<Xnnn-BOOLEAN>
             3
    ______________________________________

    ______________________________________
              (conduct <n-.vertline.test.vertline.-003>)
                 50
              (conduct <n-.vertline.test.vertline.-002>)
                 40
    ______________________________________