JP2003330940A - Creation method for retrieval index data, creation device for retrieval index data, and file retrieval device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve troubles in managing a large amount of file by classifying it according to its contents. <P>SOLUTION: A file retrieval server 100 is provided with a word extract part 102 extracting a retrieval work included in a file to be an object of a processing and an item specification part 104 specifying a retrieval item expressing the utilization of the extracted retrieval word. An index storage part 44 retains index data created in a form of corresponding the retrieval word to the retrieval item. A comparison processing part 32 acquires a target index data showing the retrieval conditions and compares the similarity between the target index data and the index data of a plurality of files. A file extract part 116 extracts a file having the contents conceptually resembling to the retrieval conditions from a plurality of files. A result provision part 118 presents the extracted file. <P>COPYRIGHT: (C)2004,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a search index data generation method and device, and a file search device that can utilize them. The present invention particularly relates to a technique for efficiently extracting a target file from a large number of files.

[0002]

2. Description of the Related Art In recent years, a PC (personal computer)
With the spread of, all documents are being digitized. Files are created electronically using document creation software such as a word processor, and these are stored in the hard disk. In an enterprise environment, computers are often connected to each other via a network, and a large number of document files are often shared by multiple users. Now that the Internet has spread, the amount of data received from outside such as web pages and emails is increasing. A predetermined administrator may classify the files in advance so that anyone can search for a desired file from such a large number of files.

[0003]

As a method of classifying a plurality of files, there is a method of grouping according to the contents of the files and storing each in a separate folder. However, it is not only difficult to uniquely define a group of files, but even if a large number of files are classified in advance by a particular administrator, the criteria for classification depend on the subjectivity of the administrator, making it difficult to search. In some cases. Thus, it is not uncommon for valuable material to remain asleep without being reused because it is both difficult to manage and retrieve.

On the other hand, some files shared by a plurality of users have a high utility value for many users, and some have a utility value for only a few users in the first place. It also includes those that are frequently reused and those that are not. Therefore, it is likely that most of the large number of files will not be needed by each user. Classification of files by type in a state where such files are mixed does not necessarily lead to easy search. If you can easily find only the files that are really useful, many users will be able to work more efficiently.

The present inventor has made the present invention based on the above recognition, and an object thereof is to provide a technique for efficiently extracting a target file from a large number of files by a highly convenient method. Still another object of the present invention is to provide a technique for accurately generating index data indicating the concept of a file to be processed.

[0006]

One aspect of the present invention relates to a method of generating search index data. In this method, when referring to a word or phrase contained in the file to be searched as a search word or phrase, the purpose of the search word or phrase is determined prior to comparison with the search condition in order to determine the similarity of the word or phrase according to the purpose. The expressed search item is specified, and the search term is held in association with the search item. Index data may be generated in such a manner that search terms are associated with search items.

A file is a document file, HTML or X
It includes files of various formats such as files generated using a description language such as ML and image files. The file mentioned here may be a record including a plurality of items, or may be a group of data including attributes of a certain entity. An attribute may be a concrete property of an entity. The search item may be an attribute of an item or entity in the record. The file may be a natural sentence, or may be in a format in which search terms are input in association with preset search items. Also,
The file may be a combination of a natural sentence and a format in which search terms are input in association with preset search items.

Another aspect of the present invention relates to a method for generating index data of a file to be searched by statistical processing based on the appearance frequency of character strings contained in the file. In this method, even if the character strings that represent the same concept are used for different purposes, the character strings are statistically processed as different character strings.

Another aspect of the present invention relates to a search index data generating apparatus. This device is a word extraction unit that extracts words and phrases included in a file to be processed, and when referring to the extracted words and phrases as search words and phrases, in order to determine the similarity of words and phrases according to the use, An item specifying unit that specifies a search item that expresses the use of the phrase and an index data holding unit that holds the search phrase in association with the search item are provided.

Another aspect of the present invention also relates to an apparatus for generating search index data. This device, in order to determine the similarity of the search term referred to at the time of search, depending on the use,
A word acquisition unit that causes a search term that expresses the purpose of the search term to be entered in a predetermined location associated with the search term, and that acquires the entered search term by associating the search term with the search term. And an index data holding unit for holding.

Here, the word / phrase acquiring unit is a questionnaire in which search words / phrases are input in association with preset search items,
The search term may be acquired in a fill-in-the-blank format such as a brochure, a catalog, or a standard table. When a natural sentence is entered in a field in which no search item is set, the phrase acquisition unit extracts the search phrase from the natural sentence and identifies the search item expressing the usage of the extracted search phrase. Good. At this time, the phrase acquisition unit may extract the search phrase by statistical processing based on the appearance frequency of the phrase included in the natural sentence. Further, the search item may be specified by referring to another field in the same file or a search item set in advance in another file.

The apparatus may further include a statistical processing section for performing statistical processing based on the appearance frequency for each combination of the search term and the search item associated with the search term. The results of various processes may be retained. The statistical processing unit may treat search terms associated with different search items, even if they have the same expression, as different terms. In addition, the statistical processing unit may treat the search terms that are substantially synonymous in consideration of the associated search items, even if the expressions are different, as the same.

Even if the search terms are different from each other, this apparatus associates with the same search item, and when the search item is taken into consideration, a similar phrase that performs a process of equating the search terms that are substantially synonymous to each other is considered. You may further have an adjustment part. The device may further include a similar phrase storage unit for each item that holds a search phrase to be input to the search item in association with the search item, and the similar phrase adjusting unit corresponds to the same search item. When one of the attached search terms is an abstract term, the similar term storage unit is referred to and the search term is associated with the embodied term.

The apparatus may further include a setting unit for setting the substance which is the subject of the search condition, and the statistical processing unit is
When the search item associated with the search term is related to the substance, the search term may be weighted higher and statistically processed. Relevance to an entity may mean that the search item includes the entity, or that the search item is an attribute of the entity.

The apparatus may further include a setting unit for setting the substance that is the subject of the file to be processed, and the item specifying unit may specify the search item in consideration of the attribute of the substance. . An attribute may be a concrete property of an entity. The setting unit may specify the concept of the file by statistical processing based on the appearance frequency of the words included in the file, and set the substance based on the concept. The search item may be specified by referring to another preset search item in the same file or a search item included in another file in which the same entity is set.

Another aspect of the present invention relates to a file search device. This device, in a plurality of files to be searched, for each file, when referring to a phrase contained in the file as a search phrase, in order to determine the similarity of the phrase according to the purpose, A target acquisition unit that acquires a search item that expresses the purpose as index data by associating it with a search term, a condition acquisition unit that acquires target index data indicating a search condition, target index data, and index data of multiple files. Based on the similarity between index data, a file extraction unit that extracts files that are conceptually similar to the search condition from multiple files, and a result presentation that presents the extracted files And a section. The index data may be generated prior to the search, or may be generated in real time during the search. The target index data is
It may be generated by the same method as the index data to be searched.

This device may be composed of a single PC, or may be composed of a system in which a server and a user terminal connected to each other via a network are combined.
In the latter case, each functional block that should be included in this device
It may be included in any of the server and the user terminal constituting the system. For example, the target acquisition unit, the condition acquisition unit,
The file extraction unit and the result presentation unit may be included in either or both of the server and the user terminal, and may be expressed by the same name in any case. When these functions are provided in the form of software modules, they may be executed by either the server or the user terminal.

It should be noted that any combination of the above components, and the components and expressions of the present invention can be realized by a method, apparatus, system,
Computer programs, recording media that store computer programs, and the like, which are mutually replaced, are also
It is effective as an aspect of the present invention.

[0019]

BEST MODE FOR CARRYING OUT THE INVENTION (Prerequisite Technology) A file search device according to this prerequisite technology searches a plurality of files for a file similar to a sentence designated by a user as a search condition. As a result, it becomes easy to search for files without classifying the files according to the contents in advance, and the burden of managing a large number of files is reduced.

FIG. 1 is a functional block diagram showing the configuration of a file search device in the base technology. The file search device 10 includes a processing unit 20 that performs a process required to generate index data that is referred to when a desired file is searched from a plurality of files, and a search process that performs a search process based on a condition designated by a user. The unit 30 and a plurality of files to be searched (hereinafter referred to as "accumulation files")
Holding unit 40 for holding data necessary for search processing
And an input / output processing unit 50 that processes input / output of data between this device and the outside.

The file search apparatus 10 can be realized by hardware such as a CPU and a memory of a computer, and can be realized by software such as a program having a file management or file search function. It depicts the functional blocks that are realized through these collaborations. Therefore, these functional blocks can be realized in various ways depending on the combination of hardware and software.

The processing unit 20 processes a plurality of accumulated files held by the holding unit 40 and extracts a plurality of characteristic character strings from each of them. The plurality of characteristic character strings form a concept (concept) that briefly shows the contents of the storage file, and the concept is recorded as index data. The processing unit 20 includes an analysis processing unit 22 that linguistically analyzes a character string included in the accumulated file, and a generation processing unit 24 that generates index data based on the analysis result.

The analysis processing section 22 includes a preprocessing section 26 and a character string extraction section 27. The preprocessing unit 26 performs preprocessing prior to language analysis. For example, a file format or a document format may be detected from the storage file to be processed, and the storage file may be converted into an atypical format such as a text format based on the file format or the document format to form an easily analyzed state. One storage file may be divided into a plurality of blocks to form a state suitable for analysis. At this time, morphological analysis, syntactic analysis,
A technique such as semantic analysis may be used.

The character string extraction unit 27 extracts a plurality of character strings from the storage file to be processed. The words included in a word dictionary described later may be extracted, or a character string delimited by spaces or blanks may be recognized as a word.

The generation processing unit 24 includes a statistical processing unit 28 and an index generation unit 29. The statistical processing unit 28 counts the appearance frequency of the extracted character string in the storage file and also the appearance frequency of the character string in all the plurality of storage files held by the file storage unit 42. At this time, the similarity between the character strings is considered. For example, synonyms,
The appearance frequency of a character string is counted by absorbing the difference between a plurality of words defined as synonyms and controlled words and having similar meanings.

The index generator 29 generates index data based on the appearance frequency of the character string counted by the statistical processor 28. This index data is configured as a list in which weights are added to the extracted plurality of character strings according to their respective appearance frequencies. As for the weighting for each character string, the character string having a higher application frequency in the storage file to be processed is set to have a higher weighting, while the character string having a high appearance frequency over all the plurality of storage files held by the file storage unit 42 is set. Lower the weight. As a result, the character string unique to the accumulated file can be revealed by a statistical method. When each accumulated file is divided into a plurality of blocks by the preprocessing unit 26, index data is generated for each block.

The holding unit 40 includes a file storage unit 4
2, an index storage unit 44, a dictionary storage unit 46, and a related data storage unit 48. The file storage unit 42 holds a plurality of accumulated files. For example, a document file generated by document creation software such as a word processor, HTML (Hyper Text Markup Language) or XML.
It includes files in various formats, such as files generated using a description language such as (eXtensible Markup Language), and the contents thereof need not necessarily be sentences. Further, the accumulated file itself does not need to be classified and standardized in advance for retrieval.

The index storage unit 44 holds the index data generated by the processing unit 20 in a form associated with the storage file. The dictionary storage unit 46 is a word dictionary,
Data such as a synonym dictionary, a synonym dictionary, and a controlled word dictionary that are referred to in the language analysis and statistical processing by the processing unit 20 are held. The related data storage unit 48 holds data that is optionally used in the processing by the search unit 30. For example, it holds a related dictionary that is referred to in order to replace a word specified as a search condition with a word of a superordinate concept, a word of a subordinate concept, a word having relevance, or the like. Such data may be generated by the processing unit 20 by extracting it from the accumulated file.

The search unit 30 receives a search condition from the user and extracts a storage file matching the search condition from the file storage section 42. The search unit 30 includes a comparison processing unit 32 that compares the search condition with the index data, and a result processing unit 34 that presents to the user an accumulation file that matches the search condition based on the comparison result.

The comparison processing section 32 includes a condition setting section 36 and a similarity determination section 37. The condition setting unit 36 acquires the search condition from the user. This search condition may be in the form of a sentence described by a natural sentence or in the form of a file containing some character string. The search condition is sent to the processing unit 20 and is subjected to the same processing as the above-described index data generation process, and the concept of the search condition is generated.

The similarity determination unit 37 detects the similarity between the search condition and the accumulated file by comparing the concept of the search condition and the concepts recorded as index data. At the time of comparison, based on various dictionaries stored in the dictionary storage unit 46 or the related data storage unit 48, another character string related to the character string included in the search condition may be added to complement the search condition. .

Here, a vector space model is used to compare the search condition with the index data. That is, the concept of search conditions and the concept of index data are expressed as vectors in a multidimensional space, and these are compared. When the concept includes n character strings, an n-dimensional vector space is formed and each component is weighted according to the appearance frequency of each character string. The degree of approximation between the vectors thus formed is the degree of similarity between the search condition and the accumulated file.

The result processing section 34 includes a list generation section 38 and a display processing section 39. The list generation unit 38 generates a list of accumulated files in descending order of similarity. At this time, the number of stored files included in the list may be adjusted to an appropriate number.

The display processing unit 39 displays a list of accumulated files as a search result on the screen. The list of accumulated files may consist of file names and a summary of their contents.

The input / output processing unit 50 is used by the file search device 1
This is an interface for inputting and outputting data between the file search device 10 and the outside thereof, such as various processing instructions for 0, input of search conditions, output of search results, and the like. When the file search device 10 is realized as a stand-alone, it serves as an interface connecting a user and this device, and when the file search device 10 is realized as a network server, communication for connecting this device with a client terminal via a network. Become an interface.

FIG. 2 is a flowchart showing the index data generation process in the base technology. First, a storage file to be processed is set from a plurality of files (S1
0), preprocesses the accumulated file (S12), and extracts a character string from the accumulated file by a process such as morphological analysis (S14). Statistical data such as the appearance frequency is calculated for each extracted character string (S16), and index data is generated based on this (S18). The storage file that has not been generated as index data is the file storage unit 42.
If the remaining files remain (S20Y), the remaining files are subjected to the processes of S10 to S18, and this is repeated until all the accumulated files are processed (S20).

FIG. 3 is a flowchart showing the search process in the base technology. First, when the user specifies a sentence as a search condition in the form of a natural sentence (S30), the processing unit 20 extracts a character string from the search condition and generates index data (S32). The index data and the index storage unit 4
4 compares a plurality of index data held therein to determine the degree of similarity (S34), generates a list of accumulated files in the order of the degree of similarity (S36), and displays this as a search result on the screen (S38). ).

In comparison with the above base technology,
An embodiment will be described. In addition, the same names and reference numerals are given to the functional blocks having the same functions as the functional blocks included in the base technology, and the description thereof will be appropriately omitted.

In the following embodiments, similar to the character string described in the above-mentioned base technology, the search word is extracted from the accumulated file to be searched or the file indicating the search condition to generate the index data. At that time, the extracted search terms are associated with the search items expressing their usage,
Index data is generated in consideration of a combination of a search word and a search item. By handling search terms in correspondence with search items, for example, it is possible to distinguish words that have different uses even if they are the same, or to identify words that are substantially synonymous with each other even if they are different words. As a result, the target file can be extracted accurately.

(Embodiment) In the present embodiment,
Processing is performed for a file in a format in which a search phrase is input in a predetermined location associated with a preset search item.

FIG. 4 is a functional block diagram showing the overall configuration of the search system in this embodiment. In the search system 80, the file search server 100 is connected to a plurality of user terminals 92 via the network 90.
The file search server 100 holds a plurality of accumulated files to be searched. The user terminal 92 is an information processing device such as a PC. The network 90 is, for example, the Internet.

FIG. 5 is a functional block diagram showing the structure of the file search server. The file search server 100 is
Similar to the search device 10 described in the base technology, it has a processing unit 20, a search unit 30, a holding unit 40, and an input / output processing unit 50. The input / output processing unit 50 is
Data is transmitted and received to and from the user terminal 92 via the network 90.

The processing unit 20 includes an analysis processing section 22 and a generation processing section 24. In the present embodiment, the analysis processing unit 22 includes a phrase extraction unit 102, an item identification unit 104, and a correspondence list generation unit 106 instead of the character string extraction unit 27 of the base technology.

The word / phrase extracting unit 102 extracts a search word / phrase included in each file. The item specifying unit 104 specifies a search item expressing the usage of each extracted search term. Further, the item specifying unit 104 adjusts the expression of the search items, for example, unifying the terms of the search items that are substantially synonymous with each other even if the phrases are different. The item specifying unit 104 may adjust the expression of the search item in consideration of the subject of the file to be processed. The correspondence list generation unit 106 generates a correspondence list in which search terms are associated with the search item.

The generation processing unit 24 includes the similar word adjustment unit 1 in addition to the statistical processing unit 28 and the index generation unit 29 of the base technology.
Including 08. The similar word adjusting unit 108 performs a process of identifying a plurality of search terms that are associated with the same search item even if the search terms have different expressions, and that are substantially synonymous in consideration of the search item. Similar word adjusting unit 108
For example, if one of the search terms associated with the same search item is an abstract term, consider all the terms that embody the search term that should be input for the search item, and Match search terms.

In the present embodiment, the statistical processing unit 28
Counts the appearance frequency for each combination of the extracted search term and the search item associated therewith. The index generating unit 29 generates index data by associating a search term with the search item. The index data is configured as a list in which a plurality of combinations of the extracted search terms and their search items are weighted according to their respective appearance frequencies. Weighting may also be done considering the subject matter of the file and the search terms. For example, search terms associated with search terms related to the subject of the file may be weighted higher. By weighting in this way, it is possible to increase the weight of important words and lower the weight of unimportant words, even for the same words, and to obtain index data that accurately indicates the concept of the file. Can be generated.

The holding unit 40 includes the correspondence list storage unit 11 in addition to the file storage unit 42, the index storage unit 44, the dictionary storage unit 46, and the related data storage unit 48 of the base technology.
0, item candidate storage unit 112 and similar phrase storage unit 114
including.

The correspondence list storage unit 110 holds the correspondence list generated by the correspondence list generation unit 106 in association with the accumulated file. Item candidate storage unit 11
The item 2 holds candidates that can be search items in association with the substance in consideration of the substance that is the subject of the accumulated file. The search item may be an attribute of the entity of the accumulated file. Further, the item candidate storage unit 112 holds a plurality of search items that are substantially synonymous with each other even if the expressions are different, in association with each other.

The item specifying unit 104 includes an item candidate storage unit 11
You may specify a search item with reference to 2. Further, when the item specifying unit 104 specifies a search item that is not stored in the item candidate storage unit 112, the item specifying unit 104 may store the search item in association with the substance. In this manner, by sequentially storing new search items in association with the entity, subsequent search items can be easily specified.

The similar phrase storage unit 114, even if the search phrase has a different expression, in combination with the search item,
Considering the search item, a plurality of search terms that are substantially synonymous with each other are held in association with the search item. For example, when the search item is "age", when a specific numerical value is input as the search term, "young", "middle-aged", and "old"
There is a case where an abstract word such as is input. The similar word storage unit 114 stores abstract terms such as “young”,
A term that embodies a numerical value such as “15 years old” is held in association with each other. When the embodied term is a numerical value as in this example, for example, the similar word storage unit 114 holds the intermediate numerical value as a reference, and the similar word adjustment unit 108
The similarity of the search term may be determined by comparison with a reference numerical value.

The search unit 30 includes a file extraction unit 116 and a result presentation unit 118 in addition to the comparison processing unit 32 and the result processing unit 34 of the base technology. In the present embodiment, the condition setting unit 36 of the comparison processing unit 32 acquires the search condition in a format in which the search phrase is input in a predetermined location associated with a preset search item, such as a questionnaire. To do. The file indicating the search condition is the processing unit 2
0, and the index data of the search condition is generated by the same processing as the above-mentioned storage file.

The file extraction unit 116 compares the index data of the file indicating the search condition with the index data of the accumulated file, and selects the search condition from the plurality of files based on the similarity between the index data. Extract files that are conceptually similar in content. The result presentation unit 118 presents the extracted file.

The accumulated file and the file indicating the search condition may be provided with a field for inputting a natural sentence that is not associated with a predetermined search item, such as a remarks field and other fields. In this case, the phrase extraction unit 102 extracts the search phrase from the natural sentence. The search term may be extracted in the same manner as the character string described in the base technology. Item identification unit 104
Specifies a search item indicating the use of the extracted search term. The item specifying unit 104 may specify the search item by referring to the item candidate storage unit 112, or may specify the search item by using a technique such as morphological analysis, syntactic analysis, and semantic analysis.

FIG. 6 is a diagram showing an example of the internal configuration of the item candidate storage unit 112. The subject of this file is vehicle product information, and the substance is "car". Item candidate storage unit 11
The item 2 holds “product name”, “maker”, “country of production”, “color of car” and the like as candidates for search items in association with the substance. Further, for example, "vehicle type" is held as a search item synonymous with "product name", and "vehicle body color" is held as a search item synonymous with "car color". In this way, by storing a plurality of search items that have substantially the same meaning even if the expressions are different in association with each other, comparison between atypical files can be easily performed.

FIG. 7 is a diagram showing an example of a storage file to be searched. The stored files to be searched are, for example, product catalogs and pamphlets. Here, this accumulated file is displayed on the user terminal 92 as a car product guide screen 130. The screen 130 shows the vehicle type column 13
2, a country of origin column 134, a manufacturer column 136, a body type column 138, a capacity column 140, a vehicle body color column 142, a price column 144 and a remarks column 146. Phrase extractor 1
02 is a search term "B234", "Germany",
"B company", "sedan", "5 people", "red", "3 million yen", etc. are extracted.

Further, the word / phrase extracting unit 102 uses the remarks column 146.
"Sporty", "red", "logo mark", etc. are extracted as search terms from the natural sentence written in.
The item specifying unit 104 uses “vehicle type” as a search item for “B234”, “country of origin” as a search item for “Germany”,
"Company" as the search item for "Company B", "Body type" as the "Sedan" search item, "5 people" as the search item for "Capacity", "Body color" as the search item for "Red", "3 million""Price" is specified as a search item for "yen". Here, since no search item is set in the remarks column 146, the item specifying unit 104 sets the item candidate storage unit 112.
For example, specify "car shape" as the "sporty" search item, "logo mark color" as the "red" search item, and "car pattern" as the "logo mark" search item. To do.

FIG. 8 is a diagram showing a correspondence list generated from the target file shown in FIG. Correspondence list 150
Is associated with each search term. Here, for example, the search term "red" is associated with "body color" and "logo mark" as search items. At this time, since these search terms “red” are associated with different search items, even the same term is treated differently.

FIG. 9 shows a screen for inputting search conditions when searching a stored file. This search screen 160 is
It allows the user to input the conditions of the desired car. The search screen 160 includes a subject column 162, a country of origin column 164, a manufacturer column 166, a body type column 168, and a vehicle body color column 17.
0, the capacity column 172, and the other column 174.

Here, the user does not enter the production country column 164 and the manufacturer column 166, and the body type column 168.
"Sedan or Light", "Red" in the vehicle body color column 170, and "Normal" in the capacity column 172. Word extraction unit 10
2 refers to the word dictionary of the dictionary storage unit 46, for example, as search terms "sedan", "light", "red", "normal",
"Sport type" etc. are extracted. Item identification unit 104
Specifies "body type" as the search items for "sedan" and "light", "body color" as the search item for "red", and "capacity" as the search item for "normal".
At this time, the search term “normal” is associated with the search item “capacity”. For example, "5 people" as a car capacity
Or “6 people” is normal, the similar word adjusting unit 108
May perform a process of equating the search term “normal” with the search items “5 people” and “6 people” associated with the same search item “capacity”.

Further, since no search item is set in the other field 174, the item specifying unit 104 refers to the item candidate storage unit 112, etc., and selects "car shape" as a search item for "sport type". Identify.

FIG. 10 is a diagram showing a correspondence list generated from the search condition file shown in FIG. In the correspondence list 180, each search term is associated with that search item. Here, the search items "country of production" and "maker" are associated with "all" as a search term. The similarity determination unit 37 of the comparison processing unit 32 matches the search condition regardless of the search terms associated with the search items “country of origin” and “maker” in the index data of the accumulated file. Then you can judge.

FIG. 11 is a flow chart showing a process of generating index data from a file to be processed. When a file to be processed is set from a plurality of files (S110), the phrase extraction unit 102 extracts a search phrase from the files (S112). Item identification part 1
04 identifies a search item for each of the extracted search terms (S114), and refers to the item candidate storage unit 112 to adjust the expression of the search item (S116). The correspondence list generation unit 106 generates a correspondence list (S118).

The similar word adjustment unit 108 performs an adjustment process between the search words that are substantially synonymous (S120), and the statistical processing unit 28 considers the search items associated with each search word while The frequency of appearance of the search term is counted (S12
2). The index generator 29 generates index data based on the frequency of appearance of the search term (S124).

The present invention has been described above based on the embodiments. It should be understood by those skilled in the art that this embodiment is merely an example, and that various modifications can be made to the combinations of the respective constituent elements and the respective processing processes, and such modifications are also within the scope of the present invention. . Hereinafter, modified examples will be described.

In the embodiment, the processing for the file in the format in which the search word is input in the predetermined location associated with the preset search item has been described. The index data may be generated and searched for as a target file. In this case, the file search server 100 may include a setting unit that sets the subject entity of the file to be processed. The setting unit may allow the user to set the substance, for example. In addition, the setting unit may generate index data indicating the concept of the natural sentence by using a prerequisite technique for extracting a character string from the natural sentence, and specify the substance of the file based on the index data. At this time, the item identification unit 104 may refer to the item candidate storage unit 112 to identify the search item.
Further, the item specifying unit 104 also refers to the search item included in another file in which the same entity is set or the search item included in another file in which the same entity is set. May be specified.

As another example, the file extraction unit 116
Based on the index data generated by the statistical processing based on the appearance frequency for each combination with the search item described in the embodiment, and based on the appearance frequency in character string units without considering the search item, as described in the base technology. A file similar to the search condition may be extracted based on both the index data generated by the statistical processing.

For example, in the embodiment, even for a search term extracted from a natural sentence, a process of specifying a search item and associating the search item with the search item was performed. You may perform the process which considered only a search term, without specifying. In the case where one file contains a part in which a search phrase is input at a predetermined position associated with a preset search item and a part in which a sentence is input by a natural sentence, each search phrase is included in the former part. Processing may be performed in association with the search item, and the latter part may be processed in consideration of only the search term as described in the base technology. In that case, the index data generated by the process of associating each search term with the search item is compared with the index data similarly generated in other files, and the index data generated by the process considering only the search term is The file extraction unit 116 may extract the file similar to the search condition in consideration of both comparison results by comparing with the index data similarly generated in other files.

[0068]

According to the present invention, it is possible to extract a file matching a search condition from a large number of files by a highly convenient method.

[Brief description of drawings]

FIG. 1 is a functional block diagram showing a configuration of a file search device in the base technology.

FIG. 2 is a flowchart showing a process of generating index data in the base technology.

FIG. 3 is a flowchart showing a search process in the base technology.

FIG. 4 is a functional block diagram showing an overall configuration of a search system according to this embodiment.

FIG. 5 is a functional block diagram showing a configuration of a file search server according to the present embodiment.

FIG. 6 is a diagram showing an internal configuration of an item candidate storage unit.

FIG. 7 is a diagram showing an example of a storage file to be searched.

8 is a diagram showing a correspondence list generated from the target file shown in FIG.

FIG. 9 is a diagram showing a screen for inputting search conditions when searching a stored file.

10 is a diagram showing a correspondence list generated from the search condition file shown in FIG.

FIG. 11 is a flowchart showing a process of generating index data from a file to be processed.

[Explanation of symbols] 20 ... Processing unit, 22 ... Analysis processing unit, 24 ...
Generation processing unit, 26 ... Preprocessing unit, 28 ... Statistics processing unit,
29 ... Index generation unit, 30 ... Search unit, 32 ...
Comparison processing unit, 34 ... Result processing unit, 40 ... Holding unit, 42 ... File storage unit, 44 ... Index storage unit, 4
6 ... Dictionary storage unit, 48 ... Related data storage unit, 50 ...
Input / output processing unit, 80 .. Search system, 90 .. Network, 92 .. User terminal, 100 .. File search server, 102 .. Word extraction unit, 104 .. Item specifying unit, 106 .. Correspondence list generation Part 108, similar phrase adjusting unit, 110 ... Corresponding list storage unit, 112 ... Item candidate storage unit, 114 ... Similar phrase storage unit, 116 ... File extraction unit, 118 ... Result presentation unit.

Claims (10)

[Claims] 1. When referring to a word or phrase contained in a file to be searched as a search word or phrase, in order to determine the similarity of the word or phrase according to the purpose, prior to comparison with the search condition,
A method for generating search index data, characterized in that a search item expressing the use of the search phrase is specified and the search phrase is held in association with the search item. 2. A method of generating index data of a file to be searched by statistical processing based on the appearance frequency of the character string included in the file, the character string representing the same concept. Even if the usage is different, the method for generating search index data is characterized in that the statistical processing is performed by using those character strings as different character strings. 3. A phrase extraction unit for extracting a phrase included in a file to be processed, and a method for referring to the extracted phrase as a search phrase,
An item specifying unit that specifies a search item that expresses the use of the search phrase, and an index data holding unit that holds the search phrase in association with the search item, in order to determine the similarity of the phrase according to the use. An apparatus for generating search index data, comprising: 4. A search term to be referred to at the time of search is input in a predetermined place associated with a search item expressing the use of the search term in order to determine whether the term is similar to the purpose, and input. Of the searched index data, wherein the acquired phrase is acquired by associating the searched item with the searched item, and an index data holding unit that holds the searched item by associating the searched item with the searched item. Generator. 5. The statistical data processing unit further comprises a statistical processing unit that performs statistical processing based on the frequency of occurrence of each combination of the search term and the search item associated therewith, and the index data holding unit includes the statistical processing. 5. The device according to claim 3 or 4, characterized in that it holds the result of (1). 6. The statistical processing unit associates search terms that are associated with the same search item even if the search terms are different from each other and that are substantially synonymous in consideration of the search item, with each other. The apparatus according to claim 5, wherein the processing is performed by performing the processing. 7. A setting unit configured to set an entity which is a subject of a search condition, wherein the statistical processing unit, when the search item associated with the search term is related to the entity, the search term. 6. The apparatus according to claim 5, wherein the statistical processing is performed by increasing the weighting for. 8. A setting unit configured to set a subject entity of a file to be processed, wherein the item identification unit identifies the search item in consideration of an attribute of the entity. The device according to claim 3. 9. In a plurality of files to be searched, when referring to a word / phrase contained in the file as a search word / word for each file, the search is performed in order to determine the similarity of the word / phrase according to the use. A target acquisition unit that acquires, as index data, a search item that expresses the use of a word in association with the search word, a condition acquisition unit that acquires target index data indicating a search condition, the target index data, and the plurality of Comparing the index data of the file, based on the similarity between the index data,
A file search device comprising: a file extraction unit that extracts files conceptually similar in content to the search condition from the plurality of files; and a result presentation unit that presents the extracted files. . 10. A process of extracting a phrase included in a file to be processed, and a process of referring to the extracted phrase as a search phrase,
In order to determine the similarity of terms according to the use, a process of identifying a search item expressing the use of the search phrase, a process of holding the search phrase in association with the search item, and executing the computer A computer program characterized by being vulnerable.