JP2005275524A - Retrieval request understanding keyword extension system using co-occurrence verb information - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a system applicable for a system for retrieving extended data intended by a retriever by allowing the retriever to designate a keyword from a large amount of stored data group such as a database or a search engine in the Internet, and to designate a system keyword for retrieving data intended by the retriever. <P>SOLUTION: This retrieval keyword extension retrieving system is developed for reducing retrieval leakage, and for improving retrieval efficiency by analyzing a retrieval keyword inputted to a retrieval keyword, and retrieving/extracting whether or not verb words and phrases exist in a retrieval keyword, and deciding the existence of noun words and phrases being a subject to the extracted verb words and phrases, and detecting words and phrases being case word and phrase information corresponding to the verb words and phrases and noun words and phrases having such subject and predictive relations, and extending the extracted case word and phrase information as the retrieval keyword. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

The present invention can be applied to a system that retrieves data intended by a searcher by designating a keyword from a large amount of data such as a database. In addition, search engines such as the Internet search system that searches for the pages intended by the searcher by specifying the keyword, or by searching for the expanded data intended by the searcher by specifying the keyword It can be applied to systems that

In a system that searches for data intended by the searcher by specifying keywords from a large amount of data such as a conventional database, only the keywords specified by the searcher are targeted. Then, collation with character information existing in the data set is performed, and data including a completely matched character string is extracted as a search result.

Also, in recent years, as a response to the fluctuation of the keyword notation entered by the searcher, even if the keyword is written in katakana such as “Aiueo”, the hiragana character string “Aioeo” of the same character string is also set as the search target keyword. Even if the alphabetic keyword is specified as “abc” in full-width lowercase letters, the keyword entered by the searcher is expanded as a search target keyword, such as half-width lowercase letters “abc”, full-width capital letters “ABC”, and half-width uppercase letters “ABC” There was a function to do. In addition, there was a function for simply expanding and interpreting keywords that can handle the fluctuations of the kana notation such as “do” and “do” in word units.

In a search system, it is common to output data containing a completely matched character string as a search result by matching the keyword entered by the searcher with the character information existing in the search target data set. It is. In addition, as described in the background art, even when the keyword input by the searcher is simply expanded and interpreted, the keyword character string created by the expanded interpretation is checked after matching the character string of the input keyword as it is. I was just collating against.

As described above, in the search system so far, the keyword entered by the searcher is not considered at all in the meaning content, but is simply treated as a symbol string and included in the data set to be searched. It was only checked against existing character information.

Under such circumstances, if a specific search image possessed by the searcher cannot be reflected in the keyword, the search result is output that is far from the searcher's intention. As described above, in the conventional search system, in addition to the amount of knowledge possessed by the searcher, the amount of knowledge indicating how much of the search system is gained from the experience of using the search system greatly depends on the search result. There was a problem that it would end up.

In order to solve the above problems, the present invention analyzes a search keyword input to a search system, searches and extracts whether a verb phrase is present in the search keyword, and then extracts the verb phrase Determine the presence of the main part of the noun phrase, detect the phrase that becomes the phrase information corresponding to the verb phrase and the noun phrase with the main predicate relationship, and expand the extracted phrase information as the search keyword By doing so, we developed a search keyword expansion search system that reduces search omissions and improves search efficiency.

The first invention to be patented is based on a basic dictionary part that stores basic word knowledge and analysis information necessary for word notation and language analysis, and a verb phrase in the basic dictionary part. A case knowledge language dictionary that holds main predicate relationship information with the main noun word and also holds a set of case information that indicates the case of the tool, means, place, etc. for each combination of main predicate relationship information A knowledge dictionary database that performs language analysis and a logic that analyzes a search keyword input to the search system, and includes a basic dictionary portion of the knowledge dictionary database that performs the language analysis. Searches for the presence of a verb phrase, and if so, extracts the verb phrase, and there is a main noun phrase for the verb phrase extracted using the knowledge language dictionary means. A search keyword analysis unit that extracts main predicate relation information in the search keyword by extracting the noun phrase when it exists, and a search keyword input to the search system by the search keyword analysis unit As a result of the analysis, if there is a verb phrase and a noun phrase that is the main part of the verb phrase, the phrase that is the phrase information corresponding to the verb phrase and the noun phrase having the main predicate relationship is stored in the case knowledge language dictionary unit. A noun that is a main part of a verb phrase and its verb phrase via the knowledge dictionary database that performs the language analysis of the search keyword input to the search system and the search keyword analysis means If the predicate relation information in the search keyword is extracted by extracting the phrase and there is a noun phrase that is the main part of the verb phrase and the verb phrase, The phrase search means is used to detect the phrase that becomes the phrase information corresponding to the verb phrase and the noun phrase having the main predicate relationship, and the detected phrase information is expanded as a search keyword, and the input search is performed. A search request comprehension keyword expansion search system using co-starter information characterized in that search omissions are reduced by searching from keywords and expanded search keywords, thereby improving search efficiency.

The first invention comprises a knowledge dictionary database that performs language analysis, a search keyword analysis means, and a phrase information search means. The search keyword input to the search system is analyzed, and the main statement relationship is determined. Understanding search requests with co-starter information that reduces search omissions and improves search efficiency by extracting phrases that become case phrase information corresponding to verb phrases and noun phrases and expanding this case phrase information as search keywords This is a keyword expansion search system.

The second invention to receive a patent is based on a basic dictionary part that stores basic word knowledge and analysis information necessary for word notation and linguistic analysis, and a verb phrase in the basic dictionary part. A case knowledge language that retains main predicate relationship information with the main noun word and phrase, and that holds case information that is a word indicating the case of each tool, means, place, etc. for each combination of main predicate relationship information It is a knowledge dictionary database for performing language analysis characterized by having a dictionary part.

The second invention is a knowledge dictionary database that performs linguistic analysis as a first means related to a system that understands a searcher's search request input to a search system, and uses basic words as knowledge to identify words. Main dictionary relation information that holds analysis information necessary for notation and linguistic analysis, and main phrase relation information for the noun word phrase that is the main part of the verb phrase in the basic dictionary section For each combination, a dictionary having a case knowledge language dictionary unit that holds case phrase information, which is a phrase indicating the case of the tool, means, place, etc., as a set is employed.

The third invention to receive a patent is a logic for analyzing a search keyword input to a search system, and holds basic analysis information necessary for word notation and language analysis using basic words as knowledge. The dictionary part is used to search for the presence of a verb phrase in the search keyword. If there is, the verb phrase is extracted, and whether the main noun phrase exists for the extracted verb phrase. The search keyword analysis means is characterized in that the main description relation information in the search keyword is extracted by determining and extracting the noun word / phrase if it exists.

The third invention employs a search keyword analysis means for extracting main description relation information in a search keyword as a second means related to a system for understanding a searcher's search request input to the search system. The search keyword analysis means is a logic for analyzing the search keyword input to the search system, and searches for a verb phrase in the search keyword using the basic dictionary part in the first means. If it does, the verb phrase is extracted and it is determined whether there is a main noun phrase for the extracted verb phrase. If there is, the noun phrase is extracted to extract the noun phrase in the search keyword. Extract main predicate relation information.

As a result of analyzing the search keyword input to the search system by the search keyword analysis means described in the third invention, the fourth invention to be patented has a verb phrase and a noun phrase that is the main part of the verb phrase. In this case, the phrase information search means is characterized in that the phrase that becomes the phrase information corresponding to the verb phrase and the noun phrase having the main predicate relationship is detected from the case knowledge language dictionary part shown in the first invention. .

  According to the fourth aspect of the present invention, as a third means related to a system for understanding a search request of a searcher input to a search system, a case is detected from a case knowledge language dictionary section shown in the first means. Adopt phrase information retrieval means. In the case where there is a verb phrase and a noun phrase that is the main part of the verb phrase in the second means, the phrase information search means includes a verb phrase having the main predicate relationship and phrase information corresponding to the noun phrase, Is detected from the case knowledge language dictionary shown in the first means.

The fifth invention to receive a patent expands the phrase information extracted by the phrase information search means shown in the fourth invention as a search keyword, thereby reducing the search omission and thereby improving the search efficiency. This is a search request comprehension keyword expansion search system based on co-starter information characterized by the above.

According to the fifth aspect of the present invention, as the fourth means related to the system for understanding the search request of the searcher input to the search system, the phrase information extracted by the phrase information search system shown in the third means is used as the search keyword. It is a search system that reduces search omissions and improves search efficiency.

A sixth invention to be patented is a knowledge dictionary database for performing language analysis according to claim 2, a search keyword analysis means according to claim 3, and a phrase information search means according to claim 4. Search by extracting a verb phrase and a noun phrase that is the main part of the verb phrase through a search keyword analysis means and a knowledge dictionary database that performs the language analysis from a search keyword specified and input by a searcher Extracting predicate relation information in keywords, and when there is a verb phrase and a noun phrase that is the main part of the verb phrase, use the phrase information search means to support the verb phrase and noun phrase that have the main predicate relation A search keyword that detects the phrase that is the search phrase information to understand, understands and infers the search intention of the searcher from the input keyword specified by the searcher, and more accurately expresses the search intention of the searcher. A search intent inference keyword generation system characterized in that so as to generate a mode.

According to the sixth aspect of the present invention, as a preceding process of the keyword search process, the search intention of the searcher is understood and inferred from the input keyword specified by the searcher, and a search keyword that more accurately expresses the search intention of the searcher is generated. A search intention reasoning keyword generation system is embodied. As a result, if a searcher's intention is inferred, replaced with an appropriate keyword, and a search process is executed, a search with a more accurate search keyword is performed, and the hit rate can be improved.

  In the present invention, not only the input search keyword specified by the searcher for the search system, but also the phrase information that is a keyword reflecting the search intention of the searcher is automatically added and expanded. Therefore, the data search means performs a data search from the search target data set using the input search keyword specified by the searcher for the search system and the expanded search keyword based on the phrase information reflecting the searcher's search intention. The search result data is output.

As described above, from the input search keyword specified by the searcher, the phrase information inferring the searcher's intention is expanded as the search keyword, and the search process is executed by both search keywords, so the hit rate of the search is improved. And the search leak rate can be reduced.

In addition, since the present invention automatically adds from the knowledge dictionary database that performs linguistic analysis when expanding keywords that reflect the search intention of the searcher, it is possible to constantly expand the search request understanding keywords stably. There is a problem that the search results greatly depend on individual differences, such as the degree of knowledge possessed by and the degree of knowledge of the search system obtained by the experience of using the search system. Can be resolved.

Hereinafter, an embodiment of a search request understanding keyword expansion search system using co-starter information according to the present invention will be described with reference to the drawings.
In the embodiment of the present invention, a search is performed on a system that searches for data intended by a searcher by inputting a keyword from a search target data set including a plurality of data. The processing is performed. Here, since the Internet is composed of a plurality of pages, if the search target data set is replaced with the Internet, the Internet search system can be realized with the same configuration.

  FIG. 1 is a system configuration diagram of an embodiment of the present invention. First, the input search keyword 1 designated by the searcher for the search system is analyzed by the search keyword analysis means 21 in the search keyword expansion system 2. The analysis procedure uses the basic dictionary unit 31 in the case knowledge language dictionary unit 3 to search for a verb phrase in the search keyword. If it exists, it extracts the verb phrase, determines whether there is a main noun phrase for the extracted verb phrase, and if so, extracts the noun phrase in the search keyword Main description relation information 22 is extracted.

  When the predicate relationship information 22 exists in the search keyword, the searcher's deep request is added to the phrase information 32 in the case knowledge language dictionary unit 3 by the phrase information search unit 23 for the extracted main statement relationship information 22. It is determined whether there is case information indicating. If it exists, the phrase information 24, which is a keyword registered in the case information indicating the deep request of the searcher, is extracted. The search keyword expansion system 2 passes the extracted phrase information 24 to the data search means 5.

  Through the above processing, the search keyword input to the data search means 5 is based not only on the input search keyword 1 specified by the searcher for the search system but also on the phrase information 24 reflecting the search intention of the searcher. Expand search keywords are also added automatically. In addition, the data search means 5 uses the input search keyword 1 specified by the searcher for the search system and the expanded search keyword based on the phrase information 24 reflecting the searcher's search intention from the search target data set 4. Search is performed and search result data 6 is output.

As described above, the present invention automatically adds and expands the expanded search keyword based on the phrase information 24 reflecting the search intention of the searcher from the input search keyword input to the data search means 5. In addition, since the search process is executed with the input search keyword and the expanded search keyword, the hit rate of the search can be improved and the leak rate of the search can be reduced.

In addition, since the present invention automatically adds from the knowledge dictionary database that performs linguistic analysis when adding / enlarging input search keywords, it is possible to constantly expand search request understanding keywords, and the knowledge amount of the searcher In addition, it solves the problem that the search result greatly depends on the amount of knowledge of how much of the search system is obtained by the use experience of the search system.

  Hereinafter, with reference to the drawings, an embodiment of a search request understanding keyword expansion system using co-initiator information according to the present invention will be described by embodying the input search keyword 1 in the above-mentioned “Best Mode for Carrying Out the Invention”. explain.

  FIG. 2 shows an example of the case where “wine skill” 11 and three words are designated as the input search keyword 1, and the input search keyword “ It is the figure which made concrete how the analysis of "Wine skill" 11 is analyzed by the data transition.

  FIG. 3 is an example in which the configuration of a part of the content of the case phrase information 32 in the case knowledge language dictionary unit 3 is extracted.

  The three words “wine craft” 11 input as the input search keyword 1 are analyzed by the search keyword analysis means 21 in the search keyword expansion system 2. In the analysis procedure, each word of “manufacturing wine” 11 is collated with a word registered in the basic dictionary unit 31 in the case knowledge language dictionary unit 3. At this time, the search keyword analysis means 21 is preferably provided with a function capable of collating even when there is a sway of notation such as adjectives and verbs, or katakana notation and hiragana notation.

  In the search keyword analysis means 21, the part of speech of the input search keyword “wine well-made” 11 obtains information of “wine (noun)”, “good (adjective)”, and “build (verb)”.

Here, since “build (verb)” is a verb, the dependency on whether or not a noun that is the main character of this verb exists is analyzed. As a result of the dependency analysis, since “wine (noun)” is a noun that can be the main character of “build (verb)”, the search keyword analysis means 21 outputs “wine-make” 221 as main description relation information.

In the phrase information search means 23, it is confirmed from “Wine-Make” 221 whether the information of the main predicate relationship exists in the phrase information 32. As shown in FIG. 3, the structure of the phrase information 32 is a pair of nouns and verbs constituting the main predicate relationship, such as “wine (noun)” 321 and “build (verb)” 322. The structure is such that it can be easily searched once the noun and verb phrases are determined as the predicate relationship.

In the phrase information search means 23, “wine (noun)” 321 and “build (verb)” 322 shown in FIG. In the main predicate relation information, there are a location case 323, a tool case 234, a means case 325, and the like in a lower hierarchy. Here, the phrase information “wine, chateau,...” 326 registered in the location case 323 and the phrase information “wine kit, barrel,...” 327 registered in the tool case 324 are searched. A phrase such as winery, chateau, wine kit, barrel,... 241 is output as the phrase information 24.

When this system is applied in this way, as shown in the above example, in response to a deep search request from a searcher who wants to “make wine”, the searcher can use “winery” and “ Even without knowing words such as “château”, these specialized terms can be automatically expanded as search keywords, resulting in an improvement in the hit rate of the search.

The present invention can be extended not only to specify a keyword as the input search keyword 1 but also to a natural language designation search system for inputting a natural language sentence. In the processing by the search request understanding keyword expansion system using co-starter information according to the present invention shown in FIG. 2, for example, a case where a natural language sentence “I want to make wine well” is input as an input search keyword 1 is taken as an example. Explain.

  The input “I want to make wine well” is analyzed by the search keyword analysis means 21 in the search keyword expansion system 2. In this case, if a morphological analysis system that divides an existing sentence into word units and determines the part of speech of the divided word is introduced into the search keyword analysis means 21, the input sentence “I want to make wine well” to “Wine “Noun”, “Good (adjective)”, and “Form (verb)” which is the final form of the verb can be extracted.

  As described above, if “wine (noun)”, “advance (adjective)”, and “build (verb)” can be extracted by introducing a morphological analysis system into the search keyword analysis means 21, the subsequent processing is performed as described above. Similar to the procedure shown in “Example 1” to “Example 2”, keyword expansion processing can be performed.

The present invention not only expands the keywords, but also understands the search intention of the searcher, and generates a search keyword that expresses the search intention of the searcher more accurately than the input keyword specified by the searcher It can also be applied as a generation system.

For example, as an input search keyword 1, the verb “build (verb)”, such as “I want to know where wine can be made well” 12 or “wine, make, place” in the natural language sentence, the verb and the main If you can extract the noun “wine (noun)” that constitutes the predicate relationship, and “location” that is the case information that exists in the main predicate relationship information “wine-making”, the keyword specified by the searcher is discarded, It is also possible to use only the word of “wine-making” as a search keyword.

When this processing is performed, since the input search keyword 1 input by the searcher must be discarded, as shown in FIG. 4, if the data search means 5 is also incorporated in a part of the search keyword expansion system 2 of the present invention, Since the data search means 5 can be controlled in the search keyword expansion system 2, this can be realized.

As described above, it is possible to infer the searcher's intention from the input search keyword 1 specified by the searcher, replace it with an appropriate keyword, and execute the search process. It will be possible to reduce the leakage rate.

As described above, the search request understanding keyword expansion system using co-starter information according to the present invention is included in a search system that executes a search by specifying keywords from a large amount of data sets such as a database search engine and an Internet search engine. If it is incorporated, it becomes possible to realize a search system that supports an operation for a searcher to specify a keyword for reflecting a search intention and can further improve search efficiency.

It is the figure which showed an example of the search request comprehension keyword expansion system structure by the co-activation verb information in the search system of this invention. It is the figure which showed an example of the transition of the data at the time of giving a specific input search keyword with respect to the search request understanding keyword expansion system by the co-starter information of this invention shown in FIG. It is the figure which showed an example of the data registered into the case phrase information in the case knowledge language dictionary part of this invention. It is the figure which showed an example of the transition of the data at the time of applying the search request understanding keyword expansion system by co-starter information in the search system of this invention, and extending to the keyword generation system which inferred the searcher's intention.

Explanation of symbols

1: Input search keyword 11: Example of input search keyword 12: Example of input search keyword 2: Search keyword expansion system 21: Search keyword analysis means 22: Main description relation information 221: Example of main description relation information 222: Main description relation Example of information 23: Case phrase information search means 24: Case phrase information 241: Example of case phrase information 3: Case knowledge language dictionary unit 31: Basic dictionary unit 32: Case phrase information 321: Example of phrase information main part 322: An example of a phrase information predicate 323: A location case of the phrase information 324: A tool case of the phrase information 325: A means of the phrase information 326: An example of a location case registration phrase of the phrase information 327: A tool of the phrase information Example 4 of case registration phrase: Search target data set 5: Data search means 6: Search result data

Claims (6)

Basic dictionary that holds basic word knowledge and analysis information necessary for word notation and linguistic analysis, and the main predicate relationship between the noun phrase that is the main part of the verb phrase in the basic dictionary Knowledge dictionary data for linguistic analysis with case knowledge language dictionary part that retains information and holds case information that is a word indicating the case of each tool, means, place, etc. for each combination of main predicate relation information Base and
A logic for analyzing a search keyword input to a search system, and searches for a verb phrase in the search keyword using a basic dictionary part of a knowledge dictionary database that performs the language analysis. Extracts the verb phrase, determines whether there is a main noun phrase for the verb phrase extracted using the knowledge language dictionary means, and if so, extracts the noun phrase A search keyword analysis means for extracting main predicate relation information in the search keyword;
As a result of analyzing the search keyword input to the search system by the search keyword analysis means, if there is a noun phrase that is the main part of the verb phrase and the verb phrase, it corresponds to the verb phrase and the noun phrase having the main predicate relationship A phrase that becomes the phrase information to be performed, and a phrase information search means for detecting from the case knowledge language dictionary part,
The main description in the search keyword is extracted from the search keyword input to the search system by extracting the verb phrase and the noun phrase that is the main part of the verb phrase through the knowledge dictionary database that performs the language analysis and the search keyword analysis means. When the relation information is extracted and there is a verb phrase and a noun phrase that is the main part of the verb phrase, the phrase information corresponding to the noun phrase and the verb phrase having the main predicate relationship using the phrase information search means The search phrase is detected, and the detected phrase information is expanded as a search keyword, and the search omission is reduced by searching from the input search keyword and the expanded search keyword, thereby improving the search efficiency. A search request comprehension keyword expansion search system using co-starter information characterized by the above. A basic dictionary that holds basic word knowledge and analysis information necessary for word notation and language analysis,
This is a phrase that holds the main predicate relation information with the noun word phrase that is the main part of the verb phrase in the basic dictionary part and indicates the case of the tool, means, place, etc. for each combination of the main predicate relation information A knowledge dictionary database for performing language analysis, comprising: a case knowledge language dictionary unit that holds case phrase information as a set. Logic that analyzes search keywords entered into the search system,
Searches for the presence of a verb phrase in the search keyword using the basic dictionary part that stores the basic word knowledge and analysis information necessary for language analysis, and if so, the verb phrase And extracting
Determine whether there is a main noun phrase for the extracted verb phrase,
A search keyword analysis means for extracting main predicate relation information in a search keyword by extracting the noun word phrase when it exists. As a result of analyzing the search keyword input to the search system by the search keyword analysis means according to claim 3, when a verb phrase and a noun phrase that is the main part of the verb phrase are present,
A phrase information retrieval means for detecting a phrase that is case phrase information corresponding to a verb phrase and a noun phrase having the main predicate relationship from the case knowledge language dictionary section shown in claim 1. By expanding the phrase information extracted by the phrase information search means according to claim 4 as a search keyword,
A search request comprehension keyword expansion search system using co-initiator information characterized by reducing search omissions and thereby improving search efficiency. A knowledge dictionary database for performing language analysis according to claim 2, a search keyword analysis means according to claim 3, and a phrase information search means according to claim 4.
By extracting a verb phrase and a noun phrase that is the main part of the verb phrase from a search keyword specified and entered by a searcher through a knowledge dictionary database that performs the language analysis and a search keyword analysis means, the main keyword in the search keyword is extracted. When predicate relation information is extracted and there is a verb phrase and a noun phrase that is the main part of the verb phrase, the phrase information corresponding to the verb phrase and the noun phrase having the main predicate relation using the phrase information search means The search keyword is detected, the searcher's search intention is understood and inferred from the input keyword specified by the searcher, and a search keyword that more accurately expresses the searcher's search intention is generated. Search intention reasoning keyword generation system.

Images