JP2006139484A - Information retrieval method, system therefor and computer program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an information retrieval method, a system therefor and a computer program, allowing retrieval of co-occurrence expression included in a desired document by the keyword retrieval in a bird's eye view without previously producing a dictionary or a pattern, without classifying documents. <P>SOLUTION: The document present in a document set database is retrieved by a query word as a retrieval keyword, a word present on the periphery of the query word is extracted in the document as a co-occurrence word, the document present in the document set database is retrieved, the number of the documents each including only the query word, the number of the documents each including only the co-occurrence word, and the number of the documents each including both the query word and the co-occurrence word are acquired, a relevance degree between the query word and the co-occurrence word is calculated on the basis of the acquired number of the documents, and a phrase and a sentence including both the query word and an important co-occurrence word are extracted from the document set database with the co-occurrence word wherein the relevance degree is present within a prescribed range as the important co-occurrence word, and are exhibited in each the important co-occurrence word as the co-occurrence expression. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention retrieves a co-occurrence expression including a co-occurrence word by extracting a co-occurrence word related to the keyword by inputting a specific keyword for a document stored on a large scale. The present invention relates to an information retrieval method and system.

  The development of the Internet has greatly increased the opportunities for individuals to send information, and users can use the Internet to examine many others' opinions on specific topics. However, when a user uses a search engine to examine a specific topic, it is difficult to assume in advance keywords that are effective for a search related to the topic, so the number of keywords that are normally input to the search engine is at most two words. In some cases, there are only 3 words. Research and development has been conducted to guide a user's search intention to be clarified by classifying and clustering search results even when inputting a small number of keywords and presenting the search results to the user [Non-patent Document 1].

  In general, document clustering is realized by assigning some weight to words included in a document, expressing the document as a vector, and grouping documents having similar vectors [Non-patent Document 2].

  There are many diary-type documents in which information is transmitted by individuals. In such documents, the ones in which the opinions and impressions of individuals are described after the date are repeated. Individual interests vary and change from day to day. As a result, there are cases where miscellaneous topics are mixed in one document. For this reason, it is difficult to classify documents in topic units in document unit classification. Moreover, since the contents of each classified document set are opened and checked, it is efficient from the viewpoint of accessing individual opinion information or grasping the trend of the entire opinion. There is also a problem that is bad.

On the other hand, when the content of the topic to be extracted is clearly determined in advance, a dictionary or pattern related to the topic is created in advance as described in [Non-Patent Document 3] and matches this. A method for extracting a part from a document set has also been proposed. However, such an approach is difficult when the content to be extracted is unknown in advance or when the topic is dynamically changed.
“Research on automatic summarization and interface for intelligent presentation of information retrieval results”, http://www.forest.eis.ynu.ac.jp/〜mori/Kaken/Informatics/ Iwanami Laboratory Software Science 15 Natural Language Processing, Shin Nagao, Chapter 11 "Opinion information extraction from Web document collection and summary generation based on focus" Tateishi et al., Proc. Of the 10th Annual Conference of the Language Processing Society (March 2004)

  As mentioned above, it is difficult to classify documents in units of topics because the document classification of search results includes various topics, and it is necessary to check the contents of classified documents individually. It is inefficient to look down.

  On the other hand, creating dictionaries and patterns related to each topic in advance also has time and cost problems.

  In order to solve this problem, the present invention makes it possible to search a co-occurrence expression included in a desired document in a bird's-eye view by searching with a search keyword without classifying the document or creating a dictionary or a pattern in advance. It is an object of the present invention to provide an information search method, a system thereof, and a computer program.

  In order to achieve the above object, the present invention provides an information search method for presenting a co-occurrence expression related to a query word as a search keyword from a document set database storing a plurality of documents using a computer device. The computer device searches for a document existing in the document set database with a query word as the search keyword, and extracts a word existing around the query word in the document as a co-occurrence word; Obtaining from the documents existing in the document database the number of documents searched only by the query word, the number of documents searched only by the co-occurrence word, and the number of documents searched by both the query word and the co-occurrence word. Calculating the relevance between the query word and the co-occurrence word based on the acquired number of documents; Collecting a co-occurrence expression that is at least one of a sentence and a phrase that simultaneously include a co-occurrence word and a query word that satisfy a predetermined condition, from the documents existing in the document set database; and the co-occurrence An information retrieval method is proposed that performs the step of presenting an expression.

  According to the information retrieval method of the present invention, a document existing in the document set database is searched with a query word, and words existing around the query word in the document are extracted as co-occurrence words. In addition, a document existing in the document set database is searched, and the number of documents including only query words, the number of documents including only co-occurrence words, and the number of documents including both query words and co-occurrence words are obtained. The degree of association between the query word and the co-occurrence word is calculated based on the acquired number of documents. Further, based on the calculated degree of association, a co-occurrence word having a strong association between the query word and the co-occurrence word is regarded as an important co-occurrence word, and at least one of a sentence and a phrase including both the query word and the important co-occurrence word One of them is extracted from a document existing in the document set database as a co-occurrence expression, and this co-occurrence expression is presented.

  Thus, for example, co-occurrence words related to a topic are extracted based on the number of matching documents from a search result with a small amount of keywords related to the topic included in the document including opinion information, and the co-occurrence of phrases and sentences including the co-occurrence word It is possible to collect expressions from a document set including opinion information, present co-occurrence expressions organized for each co-occurrence word to the user, and allow the user to search the opinion information on the desired topic from a bird's-eye view Become.

  Further, the present invention is an information search system for presenting a co-occurrence expression related to a query word as a search keyword from a document set database storing a plurality of documents in order to enable the above method. A document set database storing a plurality of documents, a search document database storing documents searched with words input from the document set database, and a co-occurrence word co-occurring with a query word in the search document The co-occurrence word list, the number of matching documents obtained when searching the document set database with the query word, the number of matching documents when searching the document set database with the co-occurrence word, and the query word. A matching document number table that stores the number of matching documents when the document set database is searched for a pair of origin words, Input a co-occurrence word list that stores important co-occurrence words and query words that satisfy a condition, a co-occurrence expression database that stores at least one of sentences and phrases that simultaneously include the query words and co-occurrence words, and a word A word input unit and a word input from the word input unit as a query word, search a document including the query word from the document set database to obtain a search document, and the search document corresponding to the query word Relevant document search unit to be stored in the search document database, each document stored in the search document database is divided into words, and words existing around the query word are used as co-occurrence words, corresponding to the query words A co-occurrence word acquisition unit for registering the co-occurrence word in the co-occurrence word list, and obtained when the document set database is searched with the query word. The number of matching documents, the number of matching documents obtained when searching the document set database with each co-occurrence word registered in the co-occurrence word list, the query word and each co-occurrence corresponding to the query word A conforming document number acquisition unit that stores in the conforming document number table the number of conforming documents obtained when the document set database is searched for a pair of words; and referring to the conforming document number table, a query word and a co-occurrence word A degree-of-association calculation unit that calculates the degree of association of the query word and the co-occurrence word satisfying a predetermined condition as an important co-occurrence word, and the important co-occurrence word corresponding to the query word An important co-occurrence word storage unit stored in the co-occurrence word list, a query word stored in the important co-occurrence word list, and each important co-occurrence word are searched for the document set database, and these words are included. A co-occurrence expression collection in which at least one of a sentence and a phrase simultaneously including a query word and each important co-occurrence word is extracted from the target document and stored in the co-occurrence expression database. And an output unit that outputs and displays the co-occurrence expressions stored in the co-occurrence expression collection unit for each important co-occurrence word.

  According to the information search method and the system of the present invention, for a topic that a user wants to examine for a document set including opinion information, a co-occurrence word that is moderately related to the topic and is not too general is based on the number of matching documents. Sentences and phrases that simultaneously contain topic words and co-occurrence words are comprehensively collected as co-occurrence expressions and presented to the user for each co-occurrence word. It becomes possible to overlook opinion information including it.

  FIG. 1 is a configuration diagram showing an information search system according to an embodiment of the present invention. In the figure, reference numeral 100 denotes an information search system, which includes a word input unit 200, a compatible document search unit 300, a document set database 400, a search document database 500, a co-occurrence word acquisition unit 600, a co-occurrence word list 700, and a compatible document number acquisition unit 800. , Conforming document number table 900, relevance calculation unit 1000, important co-occurrence word storage unit 1100, important co-occurrence word list 1200, co-occurrence expression collection unit 1300, co-occurrence expression database 1400, output unit 1500, at least It is constituted by one known computer device.

  The word input unit 200 inputs a word as a query (hereinafter simply referred to as a query word) from the outside.

  The matching document search unit 300 searches the document set database 400 for a document that matches the input query word, that is, a document including the query word, acquires a predetermined number of documents as a search document, and searches this document. The document is paired with the query word and stored in the search document database 500. Here, a large amount of documents are stored in the document set database 400 in advance.

  In the search document database 500, the search document acquired by the matching document search unit 300 is paired with the query word and stored.

  The co-occurrence word acquisition unit 600 deletes a tag from the search document stored in the search document database 500 if the document is tagged, divides each search document into words, and places the search word around the query word. Existing words are extracted as co-occurrence words, and the extracted co-occurrence words are registered in the co-occurrence word list 700 in correspondence with the query words.

  The co-occurrence word list 700 stores co-occurrence words that co-occur with query words in the search document in association with the query words.

  The number-of-matching document acquisition unit 800 is configured to obtain a document by using the number of matching documents obtained when searching the document set database 400 with each co-occurrence word stored in the co-occurrence word list 700, and a pair of the query word and each co-occurrence word. The number of conforming documents obtained when searching the collective database 400 is stored in the conforming document number table 900.

  Therefore, in the matching document number table 900, the document set database 400 is searched by the number of matching documents when the document set database 400 is searched by the co-occurrence word, and the query word and the co-occurrence word corresponding to the query word. The number of relevant documents at that time is stored.

  The relevance calculation unit 1000 refers to the matching document number table 900 and calculates the relevance of the query word and the co-occurrence word. A method for calculating the degree of association will be described later.

  The important co-occurrence word storage unit 1100 sets a co-occurrence word satisfying a predetermined degree of association between the query word and the co-occurrence word as an important co-occurrence word, and associates the important co-occurrence word with the query word as an important co-occurrence word. Store in word list 1200.

  Therefore, the important co-occurrence word list 1200 stores important co-occurrence words and query words that satisfy certain conditions in the co-occurrence words.

  The co-occurrence expression collection unit 1300 searches the document set database 400 using the query words stored in the important co-occurrence word list 1200 and each important co-occurrence word, obtains documents that match the search, and is determined in advance. For the number of documents only, sentences and phrases that simultaneously contain query words and each important co-occurrence word are comprehensively collected from the target documents, and these sentences and phrases are stored in the co-occurrence expression database 1400 as co-occurrence expressions. To do.

  Accordingly, the co-occurrence expression database 1400 stores sentences and phrases that simultaneously include the query word and the important co-occurrence word.

  The output unit 1500 outputs and displays the co-occurrence expressions stored in the co-occurrence expression collection unit 1300 for each important co-occurrence word.

  Next, the computer program processing operation of the information search system having the above-described configuration will be described with reference to the flowchart shown in FIG.

  When a word is input from the word input unit 200 (S1), the information search system 100 searches the document set database 400 for a document that matches the query word using the input word as a query word (S2), and sets the query word as the query word. The number of conforming documents is acquired and stored in the conforming document number table 900 corresponding to the query word (S3), and the number of documents within the predetermined numerical value is acquired (S4), and the retrieved search documents are acquired. Is associated with the query word and stored in the search document database 500 (S5).

  Next, if a tag is attached to the document stored in the search document database 500, the tag is deleted, each document is divided into words (S6), and words existing around the query word are shared. The extracted co-occurrence word is registered in the co-occurrence word list 700 in correspondence with the query word (S8).

  After that, the number of matching documents obtained when searching the document set database 400 with each co-occurrence word stored in the co-occurrence word list 700 is a pair of the query word and each co-occurrence word corresponding to the query word. In step S9, the number of conforming documents obtained when the document set database 400 is retrieved is stored in the conforming document number table 900.

  Next, the matching document count table 900 is referred to, and the degree of association between the query word and the co-occurrence word is calculated for each co-occurrence word (S10). A specific example of the relevance calculation method will be described later.

  Further, the information retrieval system 100 sets a co-occurrence word satisfying a predetermined degree of association between the calculated query word and the co-occurrence word as an important co-occurrence word (S11), and makes an important co-occurrence corresponding to the query word. The word is stored in the important co-occurrence word list 1200 (S12).

  Next, the document set database 400 is searched with the query words stored in the important co-occurrence word list 1200 and each important co-occurrence word (S13), and documents matching these words, that is, documents including these words are extracted. (S14) Collecting sentences and phrases including the query word and each important co-occurrence word simultaneously from only a predetermined number of documents from the target document, and co-occurring these sentences and phrases These expressions are used as expressions (S15), and these co-occurrence expressions are stored in the co-occurrence expression database 1400 (S16).

  Thereafter, the information retrieval system 100 outputs and displays the co-occurrence expressions stored in the co-occurrence expression database 1400 for each important co-occurrence word (S17).

  Hereinafter, the operation of the information search system 100 according to this embodiment will be described with reference to FIGS. 1 to 6 using a specific example.

  For example, it is assumed that the product name “digital camera 1” is input to the word input unit 200. The matching document search unit 300 sets the query word “digital camera 1” and searches the document set database 400 to extract documents that match the query word. Note that the format of the document collection database 400 here is not particularly defined, and an index as shown in [Non-Patent Document 2] may be held.

  If the number of documents set in advance is 100, for example, the conforming document search unit 300 stores the top 100 documents in the search result ranking order in the search document database 500. In this specific example, the search document database 500 includes a document ID and document text. The document ID may be a URL or the like.

  The co-occurrence word acquisition unit 600 extracts words existing in a paragraph including “digital camera 1” or the entire document from 100 documents stored in the search document database 500 and stores them in the co-occurrence word list 700. At this time, morphological analysis may be performed, for example, limiting to only words whose part of speech is a noun, including a compound word having a continuous noun, or part of speech being a verb or adjective word.

  FIG. 3 is a diagram illustrating an example of the co-occurrence word list 700. As co-occurrence words corresponding to the query word “digital camera 1”, for example, “activation”, “battery”, “small”, “light”, “image quality”, “response”, “strobe”, “shutter”, “ Stores “Manufacturer 1”, “Product”, “Digital camera 2”, and the like.

  The matching document number acquisition unit 800 acquires the number of matching documents from the document set database 400 using each of the co-occurrence words stored in the co-occurrence word list 700 as a keyword. Further, the query word stored in the co-occurrence word list 700 and each of the co-occurrence words are combined, and the number of matching documents is acquired from the document set database 400 using this as a keyword. That is, the number of documents including this keyword is acquired. For example, the number of matching documents is acquired from the document set database 400 using “digital camera 1 activation” and “digital camera 1 battery” as keywords. Next, these document numbers are stored in the matching document number table 900.

  FIG. 4 is a diagram showing an example of the conforming document number table 900. This indicates that the number of documents that match the “activation” of the co-occurrence word alone is 1,230,000, and the number of documents that match the “digital camera 1 activation” of the query word and the co-occurrence word is 1,920.

  The relevance calculation unit 1000 refers to the conformance document count table 900, calculates the relevance R (q, w) between the query word q and the co-occurrence word w by the following formula, and updates the relevance of the conformance document count table 900 To do.

  Here, H (q) is the number of documents that match the query word q with respect to the document set database 400. H (w) is the number of documents to which the co-occurrence word w matches the document set database 400. H (q, w) is the number of documents that match in the AND search of the query word q and the co-occurrence word w. Further, “*” in the formula represents multiplication.

  As shown in this equation, both the query word q and the co-occurrence word w are obtained by multiplying the number H (q) of documents including the query word q by the number H (w) of documents including the co-occurrence word w. The logarithm of the value obtained by dividing the number of included documents H (q, w) is calculated as the relevance R (q, w).

  The intent of this equation is to calculate the amount of information corresponding to the amount of mutual information in the document set of query words and co-occurrence words. For this reason, a calculation formula similar to the calculation formula for the mutual information amount is adopted here. The difference is that the mutual information amount is calculated based on the appearance frequency of two words in the document set, but it is not efficient to directly determine the appearance frequency when the document set becomes large, so here The number of documents that match two words is used.

  A co-occurrence word that is too general or too strong is estimated from the value of this calculation formula. The value of the calculation formula decreases when the co-occurrence word is too general, and increases when the co-occurrence word is strongly related. Therefore, since the important co-occurrence word storage unit 1100 needs to exclude co-occurrence words that are too general or are too strongly related to the query word as a condition for the degree of relevance R, Co-occurrence words whose degrees R (q, w) are within a predetermined range are stored in the important co-occurrence word list 1200 as important co-occurrence words. For example, the threshold range is set to -17 to -15, and the co-occurrence words having the relevance included in the threshold range are referred to the important co-occurrence word list 1200 as the important co-occurrence words with reference to the matching document number table 900. Store. For example, in FIG. 4, the co-occurrence words of “activation” having a relevance degree of −16.9 and “battery” having a relevance degree of −15.1 are stored in the important co-occurrence word list 1200 as important co-occurrence words.

  FIG. 5 is a diagram showing an example of the important co-occurrence word list 1200. The important co-occurrence word list 1200 stores “digital camera 1” as a query word and “startup”, “battery”, “small”, “light”, “image quality”, and the like as important co-occurrence words.

  Here, the reason for excluding a co-occurrence word that is too strong corresponds to the co-occurrence word “maker 1” when the query word as shown in FIG. 3 is “digital camera 1”, for example. From the information that the user wants to know, rather than the information that is known to many users, such as "Manufacturer 1" is the developer and distributor, presents useful information to the user This is because the effect is reduced.

  On the other hand, the reason for excluding too common co-occurrence words is that, for example, the co-occurrence word “product” in the case of the query word “digital camera 1” can only obtain information that is obvious to the user, and information useful for the user is also obtained. This is because it cannot be presented.

  The co-occurrence expression collection unit 1300 refers to the important co-occurrence word list 1200, creates a pair of the query word and each important co-occurrence word, for example, searches the document set database 400 using “digital camera 1 activation” as a keyword, Assuming that the predetermined number of documents is 100, the top 100 documents in the ranking order are acquired, and a part including the query word “digital camera 1” and the important co-occurrence word “activation” at the same time is extracted. At this time, the whole sentence including “digital camera 1” and “activation” may be used, a phrase having punctuation marks as a unit, or a phrase having “digital camera 1” and “activation” as both ends may be used. Locations including the query word and all important co-occurrence words are collected as co-occurrence expressions and stored in the co-occurrence expression database 1400.

  FIG. 6 is a diagram illustrating an example of the co-occurrence expression database 1400. The co-occurrence expression database 1400 also stores the document IDs of documents for which the co-occurrence expressions are collected. The document ID may be a URL or the like.

  The output unit 1500 refers to the co-occurrence expression database 1400 and outputs and displays each important co-occurrence word related to the query word and a set of co-occurrence expressions of the query word and each important co-occurrence word.

  As described above, according to the information search system of the present embodiment, a topic (query word) that a user wants to examine for a document set including opinion information is matched with a co-occurrence word that is related to the topic but is not common. Acquired based on the number of documents, collects sentences and phrases that simultaneously contain topic words and co-occurrence words as a co-occurrence expression, and presents them to the user for each co-occurrence word. It becomes possible to look down on the key keywords and opinion information including them.

  The information search system of the above embodiment is an example of the present invention, and it goes without saying that the present invention is not limited to this.

The block diagram which shows the information search system in one Embodiment of this invention The flowchart explaining the computer program operation | movement of the information search system in one Embodiment of this invention. The figure which shows an example of the co-occurrence word list in one Embodiment of this invention The figure which shows an example of the applicable document number table in one Embodiment of this invention. The figure which shows an example of the important co-occurrence word list in one Embodiment of this invention The figure which shows an example of the co-occurrence expression database in one Embodiment of this invention.

Explanation of symbols

  100: Information search system, 200: Word input unit, 300: Relevant document search unit, 400 ... Document collection database, 500 ... Search document database, 600 ... Co-occurrence word acquisition unit, 700 ... Co-occurrence word list, 800 ... Relevant document Number acquisition unit, 900 ... conforming document number table, 1000 ... relevance calculation unit, 1100 ... important co-occurrence word storage unit, 1200 ... important co-occurrence word list, 1300 ... co-occurrence expression collection unit, 1400 ... co-occurrence expression database, 1500 ... Output section.

Claims (6)

An information search method for presenting a co-occurrence expression related to a query word as a search keyword from a document set database in which a plurality of documents are stored using a computer device,
The computer device includes:
Searching a document existing in the document set database with a query word as the search keyword, and extracting a word existing around the query word in the document as a co-occurrence word;
Obtaining the number of documents searched only by the query word, the number of documents searched only by the co-occurrence word, the number of documents searched by both the query word and the co-occurrence word from documents existing in the document database; ,
Calculating a degree of association between the query word and the co-occurrence word based on the acquired number of documents;
Collecting a co-occurrence expression that is at least one of a sentence and a phrase that simultaneously include a co-occurrence word and a query word whose calculated relevance satisfies a predetermined condition from documents existing in the document set database;
And a step of presenting the co-occurrence expression. The computer device includes:
Searching a document including a query word from the document set database, and storing the acquired document in the search document database in association with the query word;
Dividing each document stored in the search document database into words, and extracting words existing around the query words as co-occurrence words;
Registering the extracted co-occurrence words in the co-occurrence word list in association with the query words;
The number of conforming documents obtained when searching the document set database with the query word, and the number of conforming documents obtained when searching the document set database with each co-occurrence word stored in the co-occurrence word list Storing in the matching document number table the number of matching documents obtained when the document set database is searched for the query word and each co-occurrence word pair corresponding to the query word;
Referring to the matching document number table, calculating a degree of association between the query word and the co-occurrence word for each co-occurrence word;
A co-occurrence word satisfying a predetermined condition for the calculated degree of association is set as an important co-occurrence word, and the important co-occurrence word is stored in an important co-occurrence word list in correspondence with the query word;
Searching the document set database for query words and each important co-occurrence word stored in the important co-occurrence word list, and extracting a document containing these words;
Extracting at least one of a sentence and a phrase including the query word and each important co-occurrence word simultaneously from the extracted document and storing it in a co-occurrence expression database as a co-occurrence expression;
The information retrieval method according to claim 1, further comprising: outputting and displaying the co-occurrence expression stored in the co-occurrence expression database for each important co-occurrence word. The computer device, when calculating the relevance, is a value obtained by multiplying the number of documents including a query word by the number of documents including a co-occurrence word, and the number of documents including both the query word and the co-occurrence word. The information search method according to claim 1, wherein a step of calculating a logarithm of a value obtained by dividing the value as the relevance is executed. An information search system that presents a co-occurrence expression related to a query word as a search keyword from a document set database in which a plurality of documents are stored,
A document set database in which multiple documents are stored;
A search document database for storing documents searched for words input from the document set database;
A co-occurrence word list that stores co-occurrence words that co-occur with query words in the search document;
The number of matching documents obtained when searching the document set database with the query word, the number of matching documents when searching the document set database with the co-occurrence word, and the document with a pair of query word and co-occurrence word A conforming document count table for storing the conforming document count when the collective database is searched;
An important co-occurrence word list storing important co-occurrence words and query words satisfying a predetermined condition among the co-occurrence words,
A co-occurrence expression database that stores at least one of a sentence and a phrase simultaneously including a query word and a co-occurrence word;
A word input unit for inputting a word;
A word input from the word input unit is used as a query word, a document including the query word is searched from the document collection database to obtain a search document, and the search document is stored in the search document database corresponding to the query word. A conforming document search unit to
Each document stored in the search document database is divided into words, words existing around the query word are set as co-occurrence words, and the co-occurrence words are registered in the co-occurrence word list in correspondence with the query words. A co-occurrence word acquisition unit,
The number of conforming documents obtained when searching the document set database with the query word, and the number of conforming documents obtained when searching the document set database with each co-occurrence word registered in the co-occurrence word list A number of conforming documents obtained by storing the number of conforming documents obtained when searching the document set database with the query word and each co-occurrence word corresponding to the query word in the conforming document number table;
A relevance calculating unit that calculates the relevance of the query word and the co-occurrence word with reference to the matching document number table;
An important co-occurrence word storage that stores an important co-occurrence word in the important co-occurrence word list in association with the query word as an important co-occurrence word with a degree of association between the query word and the co-occurrence word satisfying a predetermined condition And
The document database is searched with the query words and each important co-occurrence word stored in the important co-occurrence word list, a document including these words is obtained, and the sentence including the query word and each important co-occurrence word at the same time And a co-occurrence expression collection unit that extracts at least one of the phrases from the target document and stores it in the co-occurrence expression database as a co-occurrence expression,
An information retrieval system comprising: an output unit that outputs and displays the co-occurrence expression stored in the co-occurrence expression collection unit for each important co-occurrence word. The relevance calculator calculates a logarithm of a value obtained by dividing the number of documents including both the query word and the co-occurrence word by a value obtained by multiplying the number of documents including the query word and the number of documents including the co-occurrence word. The information search system according to claim 3, further comprising means for calculating the degree of association. A computer program in an information search system comprising a computer device that presents a co-occurrence expression related to a query word as a search keyword from a document set database storing a plurality of documents,
A computer program comprising the processing steps according to any one of claims 1 to 3.

Images