Flexible keyboard searching6321226Abstract A search engine implements a multi-level search scheme. A first level involves performing a keyword search based on character matching. A second level, performed only if the first level yields no results, is a keyword search based on phonetic representations of a search phrase and of the keywords. A third level, performed only if the first and second levels yield no results, is a rough matching search. The keywords or keyword phrases are specified in a phrase table. Each entry of the phrase table specifies a keyword phrase, its phonetic representation, a topic URL, and an action that is to be performed in conjunction with the topic URL. There are a plurality of defined actions, having different priorities. If multiple keyword phrases are found in the multi-level search, the one having the action with the highest priority is initiated. If there is a tie for the highest priority, the results are listed in a results page, regardless of the actions associated with the matched entries. Different actions can be specified in the phrase table entries, corresponding to different levels of the multi-level search scheme that were required to discover a matching entry. Claims What is claimed is: Description TECHNICAL FIELD
TABLE 1
Keyword Phonetic Topic
Phrase Spelling URL Action
Mustang mstng m.htm 1
VW vw vw.htm 1
Ford frd fd.htm 1
Lumina lmn lm.htm 1
Pony Car pncr m.htm 2
Bucket bktsts bk.htm 2
Seats
The keyword phrases are generated by a human content editor for various available topics. Thus, each topic is associated with one or more keyword phrases. The keyword phrases consist of key concepts or items described by the associated topics. As an example, a topic describing a "Toyota Tercel" might have two keyword phrases: "Toyota" and "Tercel". In some cases, it might be desirable to also include the combined words "Toyota Tercel" as a keyword phrase. There are one or more entries in keyword phrase table 102 for each keyword phrase. Each entry is associated with a single keyword phrase and references a single topic. The table is searchable by the keyword phrases. In addition to a keyword phrase, each entry also contains a phonetic spelling of the keyword phrase. Although a preferred method of formulating phonetic spellings will be described in detail below, the phonetic spellings can potentially be obtained in different ways. One simple formula for converting a keyword phrase to its phonetic equivalent is to simply remove all spaces and vowels, as is illustrated in Table 1. Each entry also contains a topic ID, which in this example is a URL indicating the location of the topic resource associated with the table entry and its keyword phrase. Finally, every entry indicates a keyword action. A keyword action indicates a default step that should potentially be taken upon matching the associated keyword phrase. More specifically, each phrase table entry (and thus each keyword phrase) is associated with one or more actions selected from a plurality of predefined actions. The actions are specified as numerical codes. Each action or action code indicates a step that should be performed whenever the table entry is identified as the result of a search. For example, one action code might indicate that the topic indicated by the topic ID of the table entry should be automatically rendered, without any further user intervention. Another action might consist of simply displaying the topic ID or URL in a "result" list for examination and selection by the user. Search and Navigation Methodology FIG. 3 illustrates steps that are performed in one embodiment of the invention for finding or identifying information resources. A step 120 comprises accepting a search phrase from a user and presenting such resources to a user. The phrase might comprise a single word or a plurality of words. It might contain alphabetical and numerical characters as well as other characters such as punctuation, asterisks, ampersands, etc. Step 122 comprises matching the search phrase with one or more keyword phrases--searching the keyword phrase table for a keyword phrase and a corresponding entry that matches the search phrase. The keyword phrase table is normally maintained as part of a database, so that this step commonly involves querying the database with the search phrase. The search strategy involves first attempting to match the user's search phrase with a keyword phrase from the keyword phrase table. If this fails, the search phrase is converted to its phonetic equivalent and an attempt is made to match the phonetic version of the search phrase with a phonetic spelling of a keyword phrase from the keyword table. For purposes of discussion, assume that the database query of step 122 identifies one or more matching keyword phrases and table entries, indicating either a character match or a phonetic match with one or more keyword phrases. Note that it is very possible for the query to identify more than one entry from a single search phrase. One reason for this is that the same keyword phrase can be specified in more than one entry, associated with a different topic ID in each entry. Another reason is that two different keyword phrases might have the same phonetic spelling, so that a given search phrase phonetically matches two different keyword phrases. Step 124 comprises determining whether the query has indeed identified more than one matching table entry. If it has, execution proceeds with step 126, which comprises listing the topic identified in the matched table entry as a hyperlink on a "results" page, along with any other topics identified during the search. If, on the other hand, only a single entry was identified, a step 128 is performed of determining whether the action code contained in the identified table entry equals the value "1". If so, execution branches to step 130, which comprises initiating the action specified by the action code. In this example, action code 1 indicates an action of immediately and automatically rendering the topic referenced by the table entry's topic ID. Step 130 therefore comprises sending the referenced topic to the client for rendering on the client's Web browser. If the result of step 128 is false, execution branches to step 132, which is a determination of whether the action code of the single identified table entry is equal to the value "2". In this example, action code 2 indicates that the topic is to be listed as a hyperlink on a results page. If this test is true, step 126 is executed, which comprises displaying the result page. If the result of step 132 is false, the search procedure ends without returning any results to the user, but displaying some type of message indicating that the search failed. Multiple Search Levels As an improvement over the search methodology described above, a multi-level search strategy is illustrated in FIG. 4. This type of search begins with a character search 140 as a first level of searching. In the character search, a search phrase is matched, character-by-character, with one or more keyword phrases. If the character search is unsuccessful (decision block 141) a phonetic search 142 is conducted, in which a phonetic representation of the search phrase is matched with phonetic representations of one or more keyword phrases. If the phonetic search is unsuccessful (decision block 143), a "rough" search 144 is conducted. A rough search is a search other than a character or phonetic matching search. The rough search might involve several different strategies, such as finding keyword phrases that are alphabetically or phonetically similar to the search phrase, without being exact matches. The rough search always returns one or more "matching" topics. After success at any level of searching, the search engine returns the results in a step 145. FIG. 5 shows steps involved in rough search 144. An initial step 150 of the rough search comprises identifying any keyword phrases that are words, sub-strings, or sub-phrases of the search phrase. A subsequent step 151 of the rough search comprises identifying any keyword phrases that match singularized and/or pluralized phonetic forms of the individual words of the search phrase (after removing prepositions and similar "non-meaning" words from the search phrase). If steps 150 and 151 yield no results (decision 153) a step 154 is performed of identifying any keyword phrases that are alphabetically or phonetically near the search phrase. This step involves identifying the two keyword phrases that surround the search phrase in an alphabetical sorting of the phrases, and the two keyword phrases that surround the search phrase in a phonetic sorting of the phrases. The proximity search always yields at least two results. Any results found during the steps described above are returned in a step 155. As an example of step 154, assume that a keyword phrase table contains the following keyword phrases: Mazda Maxima Marquee Masthead Now assume that a user submits "Masda" as a search phrase, actually looking for information on "Mazda" automobiles. Alphabetically, the nearest words are "Marquee" (before) and "Masthead" (after), so "Mazda" would not be found. Now consider the nearest words under a phonetic representation of the keyword phrases and the search phrase. In accordance with the method described below of converting phrases to their phonetic equivalents, the keyword phrases have phonetic equivalents as follows:
Keyword Phonetic
Mazda ULH
Maxima (Mxm) UQU
Marquee (mrq) UTP
Masthead (msthd) UMGH
The phonetic representation of the search phrase "Masda" is UMH. Accordingly, the keywords on opposite sides of "Masda" using this phonetic encoding are "Masthead" (before) and "Maxima." However, in accordance with the invention, phonetic encodings having the same length are given highest priority when sorting, so that "ULH" is considered phonetically "closer" to "UMH" than is "UMGH". Accordingly, this portion of the rough search will indicate "Mazda" as being adjacent the search phrase "Masda" in a phonetic ordering. The results of both the alphabetic proximity search and the phonetic proximity search are returned in step 155. Different Keyword Actions For Different Search Levels When conducting a multi-level search such as described above, different actions can be specified for a single entry of the keyword phrase table. An example of a keyword phrase table having three different actions for each entry is shown as Table 2:
TABLE 2
Keyword Phonetic Topic Action A Action B Action C
Phrase Spelling URL (char) (phonetic) (rough)
Mustang tlhv m.htm
VW da vw.htm
Ford esg fd.htm
Lumina rtu lm.htm
Pony Car pncr m.htm
Lemon rtu no.htm 1 0 0
Bucket bohlhl bk.htm
Seats
The different actions correspond to different ways of matching the search phrase with a keyword phrase. Action A indicates a step that should be initiated if an entry is identified as a result of a character search (step 140 of FIG. 4). Action B indicates a step that should be initiated if an entry is identified as a result of a phonetic search (step 142 of FIG. 4). Action C indicates a step that should be initiated if an entry is identified as a result of a rough search (step 144 of FIG. 4). With the option of specifying three different actions, the search engine can be configured to respond differently to the same phrase table entry, depending on the "exactness" of the match. A character match is considered exact, and might often result in direct navigation to the topic indicated by the table entry. A phonetic match is considered somewhat less exact, and might or might not cause direct navigation to a topic. A rough match is relatively inexact, and would in most cases cause the corresponding topic to be listed only as a "suggested" hyperlink on a results page. These behaviors are configurable for each table entry, using the fields shown above. For simplicity in discussing remaining aspects of the invention, these alternative actions are not discussed explicitly in most of the following discussion. Rather, when an action is referred to, the action is assumed to be whichever action is appropriate for a particular entry, depending on how the entry was matched or found. Alternatively, an action that is to be performed in response to a character match search will be referred to as a "character match action." An action that is to be performed in response to a phonetic search will be referred to as a "phonetic match action." An action that is to be performed in response to a rough search will be referred to as a "rough match action." Action Definitions and Priorities In an actual embodiment of the invention, there are seven defined actions, having codes 0-6. The actions are prioritized, with action 0 having the highest priority and action 6 having the lowest priority. The priorities allow the search engine to resolve situations in which multiple table entries are identified, without always reverting to the result page in this situation. Thus, the action priorities allow the search engine to automatically display a high-priority topic as the result is of a search even when the search yields multiple topics. Specifically, if multiple table entries are found in a search, the action having the highest priority will be initiated. If there is a tie for the highest priority, all of the entries (except those with action 0) will be listed on a results page. The actions will determine the location on the results page of the different entries. Each action is a step that is to be potentially performed relative to the topic ID or URL of the table entry in which the action is set forth. Actions 1-3 all indicate a step of immediately and automatically rendering the associated topic (the topic specified by the entry's topic ID). These actions are referred to as "direct navigation" actions, since they cause direct navigation to an entry's topic. The availability of three different actions of this type, having different priorities, allows certain table entries to take precedence over others when a search identifies multiple results. If there is not a tie for the highest priority, the entry having the action with the highest priority is selected, and its topic is immediately displayed. If there is a tie for the highest priority among the actions of the matched phrase table entries, the result page is displayed rather than performing any of the indicated actions. Whenever a result page is displayed, it lists all topics found in the search (except those referenced by matched phrase table entries specifying action 0). Action 1 is reserved for use with special action 0 to provide handling for so-called "forbidden" or "no-response" words. "Lemon" is an example of such a word in the automotive environment-it would be undesirable to inadvertently match "Lemon" with an automobile having a phonetically similar name, such as "Lumina." Swear words are other examples of "no-response" words. In the described embodiment of the invention, the search engine responds to these words by rendering a pre-defined topic stating that the search was unsuccessful, and perhaps suggesting some general topics of interest (such as a table of contents). Within the data architecture described, a no-response word such as "lemon" is specified in a table entry as follows:
TABLE 3
Keyword Phonetic Topic Action A Action B Action C
Phrase Spelling URL (char) (phonetic) (rough)
Lemon lmn no.htm 1 0 0
The URL "no.htm" specifies a predefined topic indicating that the matching search was unsuccessful. If the table entry is found as the result of a character search, indicating that the user entered the actual word "lemon", action 1 is initiated--the predefined topic is rendered, to the exclusion of all other topics (only "no-response" table entries specify action 1, and action 1 has the highest priority other than action 0). If the table entry is found as the result of a phonetic search or a rough search, action 0 is initiated. Action 0 is a special action indicating that the table entry is to be hidden from the user--essentially ignored by the search engine in these special situations. An entry such as this can only be matched if the user enters the exact keyword phrase. Thus, a search phrase such as "Lemmon", will not be matched with this table entry (even though it is a phonetic match for "lemon"). Rather, the "no.htm" topic will be displayed only as a result of an exact match with a specially defined "no-response" word. Actions 4-6 indicate that their corresponding topics are to be listed on a results page. With actions 4 and 5, the topics are listed as hyperlinks in respective lists, so that a user can manually select whether to navigate to a particular topic. With action 6, the topics are listed in their entirety on the results page. Topics that are the object of action 6 are generally very short (no more than three or four lines). As an example of how these different types of actions can be used, consider the following four phrase table entries and their corresponding topics:
"Lemon" "no match" topic
"Lumina" topic for the Chevrolet Lumina
"Lumina" topic for a fictitious trim option for a Chevrolet
Celebrity
"Lemmon" topic describing a fictitious new trim option
named after actor Jack Lemmon
Actions could be assigned as follows:
TABLE 4
Keyword Action A Action B Action C
Phrase (char) (phonetic) (rough)
Lemon 1 0 0
Lumina 2 2 4
(model)
Lumina 3 4 4
(trim)
Lemmon 3 4 4
Following are some examples of search phrases and the results that they would produce based on the actions specified in Table 4.
"Lemon" go directly to the "no match" topic
"Lumina" go directly to the topic for the Chevrolet
Lumina
"Lemmon" go directly to the topic describing a fictitious
new trim option named after actor Jack
Lemmon
"Lummina" go directly to the Chevrolet Lumina topic based
on a phonetic matching
"Lemona" go directly to the Chevrolet Lumina topic based
on a phonetic matching
"Lime" no character or phonetic matches, so all three
topics (other than the "Lemon" "no match"
topic) are listed as the result of a rough match.
In other words, typing any word exactly will go a corresponding topic, except for the keyword "Lumina", which shows a preference for the Lumina model rather than the Lumina trim. Typing a phonetic match for "Lumina" will always go to the Lumina model topic since, again, the phonetic actions give the model topic priority over the trim topic. A rough match lists everything on a result page. If the designer is concerned that the Lumina trim level is too obscure using this technique, the trim entry can be promoted to the same priority as the model entry, which would result in more ties and more times when the result page is displayed. Action 0 has the highest priority. Action 1 has the second highest priority. Action 2 has the third highest priority. Action 3 has the fourth highest priority. Action 4 has the fifth highest priority. Action 5 has the sixth highest priority. Action 6 has the seventh highest priority. FIG. 6 illustrates how the actions are implemented, accounting for multiple keyword phrase matches with different actions and different action priorities. An initial step 160 comprises performing a search for matching keywords, as already described above. The search potentially returns multiple matching keyword phrases and corresponding matching phrase table entries. The terms "match" and "matching" as used herein indicate some correspondence between a search phrase and a keyword phrase, not limited to an exact character or phonetic match. Each of the matching phrase table entries specifies an action. Among the actions, one has the highest priority. It is possible that more than one of the matching entries will specify this same action, resulting in a priority tie between two or more of the matching entries. A step 162 comprises determining whether any of resulting phrase table entries indicate actions of the "direct navigation" type, which are actions 1, 2, and 3. If none of the entries indicate direct navigation, a result page is displayed in a step 164. The result page contains references to all topics found during the search. If one or more of the entries indicate direct navigation actions, a step 166 is performed, comprising determining whether there is a tie between any of the entries for an action having the highest priority. Stated alternatively, this step involves determining whether two or more of the matched keyword phrases are associated with the action having the highest priority of the actions specified by the matching entries. If there is a tie, execution branches to step 164, which comprises listing references to the matched topic resources regardless of the specified actions. If only one of the matched phrase table entries has an action with the highest priority, a step 168 is performed of rendering the topic referenced by that entry. Result Pages FIG. 7 shows an example of a result page. Such a result page is displayed either when the highest priority among the actions specified by the identified matching table entries is 4, 5, or 6, or when there is a tie for the highest priority (in which case the "direct" actions are overridden). The result page has as many as three different lists of hyperlink selections. A first of the lists, generally indicated by reference numeral 174, contains primary suggestions for topics corresponding to the user's search phrase. This list contains the topic IDs or URLs resulting from any keyword entry tables having actions 2-4. A second of the lists, generally indicated by reference numeral 176, contains secondary suggestions for topics (also referred to as "see-also" topics) corresponding to the user's search phrase. This list contains the topic IDs or URLs resulting from any keyword entry tables having action 5. A third of the lists, generally indicated by reference numeral 178, contains glossary definitions, comprising topics having keywords matching the user's search phrase. This list contains the topics corresponding to any keyword entry tables having action 6 (hyperlinks are not necessary, since the entire topic is embedded in the results page). A results page in accordance with the invention might of course be formatted in many different ways, and different topics can be categorized according to different criteria. As just one example, topics can be associated with predefined types (as will be described below), and then segregated by type within a results page. Alternatively, or in addition, the format of the results page might vary depending on how the search results were found. For example, search results found as a result of a proximity search might be presented differently than results found as the result of a character search. Multi-Table Architecture Although the architecture shown above is very useful, it requires entry of alternative actions for each different phrase table entry. An improved, multiple table architecture, eliminates this requirement. In accordance with this embodiment of the invention, there are three tables: a topic table, a keyword phrase table, and a behavior table. An example of a topic table is shown below as Table 5.
TABLE 5
Topic
Topic ID URL Topic Type
1 m.htm 1 (model)
2 vw.htm 2 (make)
3 fd.htm 2 (make)
4 lm.htm 1 (model)
5 pc.htm 3 (feature)
6 no.htm 7 (special)
7 bk.htm 6 (glossary)
The topic table has an entry (row) for each topic resource. Each entry has fields corresponding to a topic ID number, an associated topic URL, and a topic type. In the current embodiment of the invention, there are seven topic types as follows: 1. Model 2. Make 3. Feature Article 4. News Article 5. Shopping Article 6. Glossary 7. Special (for "no-response" topics) These types correspond to different types of topics available to the server. An example of a keyword phrase table in accordance with this embodiment of the invention is shown below as Table 6.
TABLE 6
Keyword Phonetic Topic
Phrase Spelling ID Category
Mustang tlhv 1 1
VW da 2 2
Ford esg 3 1
Lumina rtu 4 1
Pony Car pncr 5 2
Lemon rtu 6 1
Bucket bohlhl 7 2
Seats
Each entry of the phrase table corresponds to a keyword phrase. In addition to the keyword phrase, each entry indicates a topic ID (referencing the topic table), the phonetic spelling of the keyword phrase, and an entry categorization. An example of a behavior table is shown below as Table 7.
TABLE 7
Action A Action B Action C
Type Category (char) (phonetic) (phonetic)
1 1 1 2 2
2 0 1 2 2
3 0 1 2 2
1 2 2 0 0
4 1 2 2 2
4 2 3 3 2
5 1 2 2 2
5 2 3 3 2
6 0 4 4 2
7 1 1 0 0
The behavior table has indexed fields corresponding to topic type and entry categorization. A category value of 0 in Table 7 indicates that the corresponding entry applies to any category within the given type. For each possible combination of topic type and entry categorization (there are only two possible entry categorizations in this embodiment of the invention--1 and 2), the table indicates one or more alternative action codes, corresponding to character matches, phonetic matches, and rough matches. In accordance with this embodiment, each phrase table entry is associated with one or more actions as described in the previous embodiment. However, the actions are found by first determining the topic type of the topic indicated by the entry and determining the entry categorization of the entry. These parameters are used to index the behavior table to find the appropriate action code. This feature allows different keyword phrases to be easily categorized without specifically entering the action codes for each phrase. In a simple implementation, there is only one possible value for the entry categorization, and the appropriate actions are determined from the topic types. Thus, it is only necessary to categorize the topics by type to specify actions to be performed when individual topics are matched in a search. In a slightly more complex embodiment of the invention, there are two possible entry categorization values. This allows a content designer to provide different actions depending on which keyword phrase was used to find the topic. Overall Search and Navigation Summary The different invention aspects described above can be integrated to form an effective search and navigation strategy. FIG. 8 shows an example of such a search and navigation strategy. A step 200 comprises matching a search phrase provided by a user with one or more keyword phrases in a keyword phrase table, using a character matching search to find an exact character match for the search phrase. If step 200 yields one or more matching table entries (step 201), a step 202 is performed of determining whether any character match actions associated with the matching table entries have priorities of 1, 2, or 3. If they do not, execution branches to step 203, which comprises displaying a result page such as shown in FIG. 7. If there is a priority of 1, 2, or 3, execution proceeds from block 202 to another decision 205, which determines whether there is a tie between the matched entries for the action having the highest priority. If there is a tie, step 203 is performed. If there is not a tie, a step 206 is performed of automatically rendering the topic specified by the table entry having the action with the highest priority. If the result of step 201 is negative, indicating that the character search yielded no results, a step 207 is performed, comprising a phonetic matching search for matching keyword phrases and their keyword phrase table entries. If step 207 yields one or more matching table entries (step 208), execution branches to step 202 (which in this case will evaluate the phonetic match actions). If the result of step 208 is negative, indicating that the phonetic search yielded no results, a step 209 is performed, comprising a rough search. After the rough search, step 210 is performed of displaying the result page. As described above, the rough search will always match one or more entries, which will be listed according to their rough match actions in the result page. Generally, in compiling the result page, the search engine will organize the various matched topics according to their actions. Note that this strategy handles three different scenarios. In response to matching a search phrase to a single keyword phrase as a result of the character matching search 200 or the phonetic matching search 207, the search engine initiates the action associated with the single keyword phrase. In response to matching the search phrase to one or more keyword phrases as a result of the rough match search 209, the search engine lists references to the topic resources associated with the one or more keyword phrases for potential selection by a user, regardless of the actions associated with the multiple keyword phrases. In response to matching the search phrase to multiple keyword phrases (as a result of any type of search), the search engine lists references to the topic resources associated with the multiple keyword phrases for potential selection by a user, again regardless of the actions associated with the multiple keyword phrases. Phonetic Sorting The invention includes a method of phonetically representing search phrases for phonetic matching purposes. In accordance with this aspect of the invention, a keyword phrase table such as described above includes a phonetic representation of its search phrase. The phonetic representation is arrived at by performing the steps of FIG. 9. A first step 300 comprises removing spaces and punctuation, and converting all characters to lower case. A step 301 comprises removing any non-leading vowels ("a", "e", "i", "o", "u", and "y") from the keyword phrase--removing all vowels unless they occur as the first character in the keyword phrase. A step 302 comprises translating individual consonant groups of the keyword phrase to a simplified phonetic code having characters representing basic sounds of speech. The term "consonant group" indicates two or more consonants that are articulated as a single sound, such as "ch", "ng", "sch", etc. Step 303 comprises translating remaining consonants of the search phrase to the simplified phonetic code. Table 8 shows how translation steps 301, 302 and 303 are accomplished. Characters from the keyword string, referred to in the table as "input characters" are converted to the "phonetic characters" in the matching row of the table. Thus, the simplified phonetic code includes the characters listed under the "phonetic characters" headings of Table 8.
TABLE 8
Input Phonetic Input Phonetic
Character Character Character Character
a remove gh H
b B ck K
c K bb B
d D cc K
e remove dd D
f F ff F
g G gg G
h H hh H
i remove jj J
j J kk K
k K ll L
l L mm M
m M nn N
n N pp P
o remove rr R
p P ss S
q K tt T
r R vv V
s S ww X
t T xx Y
u remove zz Z
v V 0 0
w W 1 1
x X 2 2
y remove 3 3
z Z 4 4
ch @ 5 5
sh $ 6 6
th % 7 7
ng & 8 8
ph F 9 9
As shown in FIG. 9, the consonant groups are translated first, followed by any remaining consonants. After steps 302 and 303, the keyword string has been translated to a phonetic spelling. A further step 304 comprises mapping each character of the phonetic spelling to an alternate coding to provide a new sort order based on a linguistic pattern. Specifically, the consonant characters of the simplified phonetic code described above are sorted to start with sounds made at the front of the mouth and proceeding back. For example, "W" is a sound that is produced at the far front of the mouth. The labial stops, "B" and "P" are produced slightly farther toward the rear of the mouth. Next come the labial fricatives "V" and "F", followed by the dental fricative "Th". The dental stops "D" and "T" are next, followed by the aleodental diphthongs "J" and "Ch", the aleodental fricatives "Z", "S", and "Sh", the velar stops "G" and "K", the velar diphthong "X" and the glottal "H". The last few consonants are put at the end of the sort order, near each other: laterals "L" and "R" and nasals "M", "N", and "Ng". Vowels are located at the first of the sort order, and numbers are listed at the end of the sort order. The resulting sort order is as follows:
A E I O U W B P V F Th D T J Ch Z S Sh
G K X H L R M N Ng 0 1 2 3 4 5 6 7 8 9
To accomplish this sort order, these characters are mapped or further translated to sequential characters of the available character set. Specifically, the characters of the simplified phonetic code are remapped as follows:
TABLE 9
Phonetic Remapped Phonetic Remapped
Character Character Character Character
A ! F E
E # % (Th) F
I $ D G
O % T H
U & J I
0 0 @ (Ch) J
1 1 Z K
2 2 S L
3 3 $ (Sh) M
4 4 G N
5 5 K O
6 6 X P
7 7 H Q
8 8 L R
9 9 R S
W A M T
B B N U
P C & (Ng) V
V D
As an example, consider the keyword phrase "sierra pickup". In accordance with step 300, all spaces are removed resulting in "sierra pickup". Next, non-leading vowels are removed, leaving "srrpckp". Then, consonant groups are translated in accordance with Table 8, yielding "sRpKp". The remaining consonants are then translated, producing "SRPKP" (no changes in this example). Finally, Table 9 is used to remap these characters to "LSCOC". Note that the remapping shown above is one of several different possible remappings. Alternatively, voiced consonants can be ordered prior to their voiceless counterparts, yielding the sort order:
A E I O U W P B F V Th T D Ch J S Z Sh
K G X H L R M N Ng 0 1 2 3 4 5 6 7 8 9
When performing phonetic searches as described in preceding sections of this description, a search phrase (or its individual words-see step 152 of FIG. 5) is converted to its phonetic representation using the steps of FIG. 9. It is then compared against the phonetic representations stored in the keyword phrase table. When performing a phonetic proximity search (step 154 of FIG. 5), the keyword phrase entries are sorted according to their phonetic keyword phrase representations. The appropriate location of the search phrase within this sorted list of phrases is then determined. Immediately preceding and following phrase entries are then returned as "rough" results of the search. To improve results, phrase table entries whose phonetic representations match the length of the search phrase's phonetic representation are preferred over other entries. Conclusion The invention provides significant advantages over prior search schemes. The invention allows different responses depending on how different topics are identified. Closely matching results can be immediately rendered, while more speculative results can be presented in a results list. Furthermore, a Web site designer has great flexibility in deciding how any particular match will be handled. These features provide a much friendlier environment for users of the Web site. Although the invention has been described in language specific to structural features and/or methodological steps, it is to be understood that the invention defined in the appended claims is not necessarily limited to the specific features or steps described. Rather, the specific features and steps are disclosed as preferred forms of implementing the claimed invention.
|
Same subclass Same class Consider this |
||||||||||