JP5802924B2 - Document search system and document search program - Google Patents

Description

  The present invention relates to a document search system and a document search program that search a document database and display search results.

  Conventionally, a search system has been used to extract necessary information from a vast amount of information. In a general search system, a document containing a lot of input search terms is displayed at the top, and further, other factors are taken into consideration (see paragraph 0029 of Patent Document 1).

JP 2009-187211 A

  However, in the above display method, when the input search word includes general-purpose words and non-general-purpose words, a document containing a large number of general-purpose words is displayed at the top, and the user really needs it. The document is displayed at the bottom. In addition, in the above display method, when a plurality of search terms are used, documents that do not include sentences in which the search terms are associated with each other are simply ranked in a high rank. There is. As a result, in order to search for a document that is really necessary, the user has to display and browse the search result document sequentially, which requires a great deal of time and effort.

  However, a user who is good at searching may be able to search while avoiding general-purpose words. However, no matter how good the user is, it is difficult to know which term is a general term in the database to be searched before searching. From the above, even if the user uses a plurality of search words including general-purpose words, the advent of a search system that displays a document that the user really needs at the top is awaited.

  The present invention has been made in view of such a problem, and an object of the present invention is to reduce the influence of general-purpose words and to display a document including a sentence in which search terms are related to each other at a higher level. To provide a search system and a document search program.

In the document search system according to the present invention, the appearance frequency for each word obtained by dividing each document in the search target document group in units of n characters (n ≧ 1) is registered for each format division of each document. This is a system that uses an index. This system includes the following three components.
(A1) A dividing unit that analyzes a given search condition and divides each search word included in the search condition in units of n characters. (A2) Obtained by dividing each search word using the above index. An extraction unit that extracts the appearance frequency for each word for each format break registered in the above index (A3) The frequency of occurrence for each word extracted by the extraction unit is used, and is configured in the document of each search word. Without using location information within the format break, calculate the appearance frequency of each search term for each format break and the generality of each search word, and use the appearance frequency and versatility obtained by this calculation. A weighting unit that calculates the weight of each document

In the document search program according to the present invention, the appearance frequency for each word obtained by dividing each document in the search target document group in units of n characters (n ≧ 1) is registered for each format division of each document. This program uses an index. This program causes a computer to execute the following three steps.
(B1) First step of analyzing a given search condition and dividing each search word included in the search condition in units of n characters. (B2) Obtained by dividing each search word using the above index. and the frequency of occurrence of each word, as well as utilize the frequency of appearance of each word extracted in the second step (B3) extraction unit for extracting each form separator which is indexed above, in the document for each search term Without using the location information within the configured format break, calculate the appearance frequency of each search term for each format break and the generality of each search term, and calculate the appearance frequency and generality obtained by this calculation. Third step to calculate the weight of each document using

  In the document search system and the document search program according to the present invention, the versatility of each search term is calculated using the above-described index. As described above, in the present invention, since the versatility of each search term is derived at the time of search, ranking display in consideration of the versatility can be performed. In addition, since the general degree of each search word is derived by calculation, the user does not need to worry about whether or not the word input as the search condition is a general word. Further, in the present invention, the appearance frequency of each search word is calculated for each format break smaller than the document using the above-described index. As a result, it is possible to prevent a document in which search terms are not related to each other from being ranked high by simply making an exhaustive description.

  In the document search system and the document search program according to the present invention, the format separator is, for example, a page, a paragraph, a chapter, or a section. The document search system according to the present invention may further include a merging unit that determines the ranking of each document using the weight obtained by the weighting unit. The document search system according to the present invention may further include not only the merge unit but also a search result display unit that displays each document according to the ranking determined by the merge unit. Here, the search result display unit may display a plurality of continuous format partition layouts including the format partition having the highest appearance frequency in each document. In addition, the search result display unit may display a format-delimited layout that maximizes the appearance frequency in each document.

  In the document search system and the document search program according to the present invention, the n character unit may include a plurality of character units, and the index may include an index for each character unit included in the n character unit. In this case, the dividing unit may divide each search word in units of characters. Further, the extraction unit may extract an appearance frequency for each word obtained by division in the division unit by using an index for each format break registered in the index and for each character unit. Good. Furthermore, the weighting unit uses the appearance frequency for each word obtained by the extraction in the extraction unit to calculate the appearance frequency for each format term and each character unit, and the generality of each search word. However, the weight of each document may be calculated using the appearance frequency and the versatility obtained by this calculation.

  According to the document search system and the document search program of the present invention, the appearance frequency for each word in units of n characters obtained from the document group to be searched is registered using each index registered for each format delimiter. Calculation of the generality of the search terms and the frequency of occurrence of each search term for each format break smaller than the document reduces the influence of the general terms and includes sentences in which the search terms are related to each other Documents can be displayed at the top.

It is a functional block diagram of the document search system which concerns on one embodiment by this invention. It is a functional block diagram of the index registration part of FIG. It is a figure showing an example of an index structure. It is a figure showing an example of an index. It is a functional block diagram of the search part of FIG. It is a figure showing an example of the calculation in the search part of FIG. It is a figure showing an example of a display of a search result. It is a figure showing the other example of a display of a search result. It is a figure showing the other example of a display of a search result. It is a functional block diagram of the modification of the document search system of FIG. It is a figure showing an example of the index structure of FIG. It is a functional block diagram of the search part of FIG. It is a block diagram of the 1st application example of the document search system of FIG. It is a block diagram of the 2nd application example of the document search system of FIG. It is a block diagram of the 3rd application example of the document search system of FIG. It is a block diagram of the 4th application example of the document search system of FIG.

DESCRIPTION OF EMBODIMENTS Hereinafter, embodiments for carrying out the invention will be described in detail with reference to the drawings. The description will be given in the following order.

1. 1. Example in which a single index is used 2. Modified example An example in which a plurality of indexes are used. Application examples

<1. Embodiment>
[Constitution]
FIG. 1 shows an example of functional blocks of a document search system 100 according to an embodiment of the present invention. The document search system 100 includes, for example, a document storage unit 110, an index registration unit 120, an index 130, a search condition input unit 140, a search unit 150, a merge unit 160, and a search result display unit 170, as shown in FIG. ing.

  The document storage unit 110 stores a document group to be searched. The document storage unit 110 includes, for example, a storage device accessed via a network, such as NAS (Network Attached Storage), or a hard disk accessed via a bus. The “search target document group” mentioned above refers to a document group registered in the index 130 (that is, an address is known). The document group is a set of documents created by various editors. The documents are, for example, office documents, academic papers, periodicals, and the like.

  FIG. 2 shows an example of functional blocks of the index registration unit 120. The index registration unit 120 creates and registers an index of a document group in the document storage unit 110. The index registration unit 120 may be configured with hardware (application circuit), or may be configured with an arithmetic device loaded with a program (software).

  First, the index registration unit 120 acquires and creates a list of documents (steps S101 and S102). Specifically, the index registration unit 120 acquires, for example, file name, address, date, and file size information for each document in the document storage unit 110 and lists them. At this time, the index registration unit 120 may assign one identifier for each acquired document. The identifier at this time may be anything as long as it is unique for each document, and may be any number or symbol, for example.

  Next, the index registration unit 120 extracts an address from the created list, and acquires a document corresponding to the address from the document storage unit 110 (step S103). At this time, if the index registration unit 120 has a list created in the past, the index registration unit 120 takes a difference between the past list and the current list, and only when a new document or an updated document is detected, The document is acquired from the document storage unit 110. The update of the document can be determined from, for example, the date and the file size. Furthermore, when the index registration unit 120 detects that there is no known document when the difference between the past list and the current list is obtained, the index registration unit 120 deletes the document from the current list.

  Next, the index registration unit 120 performs page division on each acquired document (step S104). Specifically, the index registration unit 120 assigns one identifier to each page of each acquired document. The identifier at this time may be anything as long as it is unique for each page in each document, may be a simple page number, or may be any number or symbol.

  Here, the significance of performing page division will be described. In general, when a search is performed on a file basis, a document having a large file size or a document in which a wide range of information is described is superior. However, such a document is not always a document that a user really needs. In particular, when a plurality of search terms are used, there is a high possibility that the documents as described above are documents in which the search terms are not related to each other. A document in which the search terms are not related to each other is not a document that the user really needs and is not a document to be displayed at the top. On the other hand, when a search is performed in units of pages, the file size and the width of the description width do not affect the superiority of the document. Furthermore, when a plurality of search terms are used and all the search terms are distributed within one page, the search terms can be compared with each other even if the position of the search terms in the document is not grasped. Are likely to be related. Accordingly, by performing a search in units of pages, it is possible to obtain a result equivalent to a search that takes into account the position of the search term in the document.

  In order to consider the position of the search word in the document, it is necessary to search each document to be searched in the document storage unit 110 by the grep type (text total name type) for each search word. . In a grep type search, a very large load is imposed on the processing, and it is difficult to perform a high speed search. On the other hand, when a search is performed in units of pages, the position information in the document of the search term is not necessary in the first place, and therefore it is not necessary to perform a grep type search at the time of the search. Is possible.

  Next, the index registration unit 120 performs n character division (n ≧ 1) for each page of each acquired document (step S105). Specifically, the index registration unit 120 cuts out sentences with n characters for each page of each acquired document. At this time, when the number of characters in the sentence is m, the sentence is divided into (m− (n−1)) words. For example, taking a sentence consisting of 12 characters such as “Keyword is input.”, This sentence is “key”, “-wa”, “wa”, “do”, “do”, [ Is input, ”“ input, ”“ strength, ”“ done, ”“ re, ”and“ ru. ”Are divided into 11 words.

  Here, the significance of performing n character division will be described. In general, as a method of creating an index, a method of creating an index for a search term prepared in advance, and a method of creating an index for a word obtained by dividing a sentence by n characters (n -Gram). The method used in the present embodiment is the latter n-gram. In the former method, it is necessary to prepare a search word in advance, so it takes time to prepare a search word in advance. On the other hand, in the case of n-gram, since a word is automatically extracted if there is a document group to be searched, it is not necessary to prepare a word in advance. Thus, by applying n-gram, the labor required for the search can be greatly reduced.

  Next, the index registration unit 120 creates a split index (step S106). Specifically, the index registration unit 120 registers the word obtained by dividing the sentence in the divided index for each page of each document, and when duplicate words are obtained, the number of occurrences of the word is calculated. Increment and register. Therefore, a word and the number of appearances are registered in the division index as a pair for each page of each document.

  Next, the index registration unit 120 merges and registers the indexes (steps S107 and S108). Specifically, for example, as illustrated in FIG. 3, the index registration unit 120 changes the divided index to a structure (index structure 121) that is finally used for search. In the index structure 121, the appearance frequency for each word obtained by dividing a sentence is associated with each page of each document. The index structure 121 includes, for example, a word (word 121A) obtained by dividing a sentence, an identifier of a document including the word (file number 121B), an identifier of a page including the word (page number 121C), and the word The number of appearances in one page (appearance frequency 121D) is a paired structure. For example, as illustrated in FIG. 4, the index registration unit 120 creates an index structure 121 for each word obtained by dividing a sentence and for each page of each document, and registers the index structure 121 in the index 130.

  The search condition input unit 140 receives a search condition input by the user. The search condition input unit 140 may be a data input device such as a keyboard, a mouse, a touch panel, and a microphone, or may be a communication device that receives a search condition input by a user via a network, for example.

  FIG. 5 illustrates an example of functional blocks of the search unit 150. The search unit 150 extracts a document that meets the search condition input from the search condition input unit 140 from a search target document group in the document storage unit 110 based on the index 130. The search unit 150 may be configured by hardware (application circuit), or may be configured by an arithmetic device loaded with a program (software).

  First, the search unit 150 analyzes the search condition given from the search condition input unit 140 and extracts a search word (keyword) included in the search condition (step S201). At this time, the search condition may include only one search term, but generally includes a plurality of search terms. In the following description, it is assumed that a plurality of search terms are included in the search condition. For example, when the input search condition is “keyword Kensaku”, the search unit 150 extracts two words “keyword” and “kensaku” as search words as shown in FIG.

  Next, the search unit 150 performs n character division for each acquired search word (step S202). Specifically, the search unit 150 cuts out each acquired search word in units of n characters. At this time, the number (n) of characters to be extracted is the same as the number (n) of characters to be extracted in the n character division performed when the index 130 is created. For example, as shown in FIG. 6, in the “keyword”, the number of characters m is 5, the number of characters to be extracted (n) is 2, and the number N of extraction is m− (n−1) = 5− (2-1) = 4. Therefore, the search unit 150 divides the “keyword” into four words “key”, “-wa”, “wa”, and “do”. Further, for example, as shown in FIG. 6, “kensaku” has a character number m of 4, a character number (n) to be cut out is 2, and the number N of cuts is m− (n−1) = 4− (2-1) = 3, the search unit 150 divides “kensaku” into three words “ken”, [nusa], and “saku”.

  Next, the search unit 150 uses the index 130 to calculate the appearance frequency of each search word for each document registered in the index 130 (step S203). Specifically, first, the search unit 150 uses the index 130 to extract the appearance frequency for each word obtained by dividing each search word for each page of each document registered in the index 130. For example, as illustrated in FIG. 6, the search unit 150 acquires 10 on page 1 of file number 5 and 4 on page 6 of file number 8 as the appearance frequency of “key”.

  Next, the search unit 150 calculates (estimates) the appearance frequency of each search word by using the appearance frequency for each word obtained by extracting each page of each document. For example, as shown in FIG. 6, in page 1 of file number 5, the appearance frequency of “key” is 10, the appearance frequency of “-wa” is 5, the appearance frequency of “war” is 8, and “do” Therefore, the search unit 150 uses the minimum value of 2 as the occurrence frequency of “keyword” in the file number 5 (or page 1 of the file number 5) (f5 (keyword)). And By using the same method, the search unit 150 sets 1 to the appearance frequency (f8 (keyword)) of the “keyword” in the file number 8. Similarly, the search unit 150 sets 13 as the appearance frequency (f5 (kensaku)) of “kensaku” in the file number 5 and 16 as the appearance frequency (f8 (kensaku)) of “kensaku” in the file number 8.

  Next, the search part 150 calculates the versatility of each search term using the index 130 (step S204). Here, the general-purpose degree means the degree of distribution in the document group to be searched in the document storage unit 110. The versatility is a concept corresponding to the number of documents (so-called hit number) or the ratio of documents including the search word in the document group to be searched in the document storage unit 110. Words with a high degree of versatility are words that are widely distributed in the search target document group in the document storage unit 110, and are words that are not sufficiently effective for narrowing down the search.

  Here, the significance of calculating the versatility of each search term will be described. In a general search system, “keyword order” is applied in which documents including many input search terms are displayed at the top. However, in such a search system, if the input search word includes general-purpose words and non-generic words, a document containing many general-purpose words is displayed at the top, and a document containing non-generic words is displayed. Is displayed at the bottom. At this time, the document that the user really needs seems to be a document containing many non-generic words, but in the above search system, such a document is obstructed by the generic words and buried in the lower level. . From this, it can be seen that it is particularly important to consider the versatility of the search term when displaying the search result, in order to display the document that the user really wants at the top.

  The versatility may be, for example, the number of hits obtained by searching a document including a search word in a document group to be searched in the document storage unit 110, and the appearance frequency of the search word is 1 or more. May be the number of documents. In the case of applying the number of documents in which the appearance frequency of the search word is 1 or more as the versatility, the search unit 150 calculates the appearance frequency of each search word in the document having the appearance frequency of 1 or more. It is possible to count the number and use the count number obtained as a result as the versatility. That is, when the number of documents in which the appearance frequency of the search word is 1 or more is applied as the general degree, the search unit 150 searches the document storage unit 110 to calculate the general degree of each search word. There is no need to search the entire target document group.

  Next, the search unit 150 performs document weighting (step S205). Specifically, the search unit 150 first lists documents including each search term. For example, the search unit 150 lists a document (or an identifier of the document) in which the appearance frequency of the search word is 1 or more, or a document (or an identifier of the document) in which the appearance frequency is 1 or more. Next, the search unit 150 calculates the weight for each search word of each document listed using the appearance frequency and the versatility of each search word.

  When the appearance frequency of each search word is fd (key), the general-purpose degree of each search word is V (key), and the number of documents in the document group to be searched in the document storage unit 110 is M, the search unit 150 For example, as shown in FIG. 6, the weight of each listed document for each search word is obtained using fd (key) × M / V (key). Further, the search unit 150 obtains the weight Cost (d) of each listed document using, for example, Σ (fd (key) × M / V (key)) as shown in FIG. In the formula, d is a file number. For example, as illustrated in FIG. 6, the search unit 150 acquires 27.6 as the weight Cost (5) of the file number 5 and uses the above formula as the weight Cost (8) of the file number 8. Get 25.2. In other words, in the example of FIG. 6, the weight Cost (5) of the file number 5 is larger than the weight Cost (8) of the file number 8 in “keyword” and “kensaku”.

  Here, as shown in FIG. 6, the appearance frequency of “kensaku” is an order of magnitude higher than the appearance frequency of “keyword”. Therefore, when the general “ranking order” is used, the evaluation value of the file number 8 should be larger than the evaluation value of the file number 5. Therefore, in this case, a document with a high occurrence frequency of “kensaku” (a document with file number 8) is displayed at the top, and a document with a high appearance frequency of “keyword” (a document with file number 5) is displayed at the bottom. The On the other hand, in the example of FIG. 6, the weight Cost (5) of the file number 5 is larger than the weight Cost (8) of the file number 8. Therefore, in the example of FIG. 6, a document with a high occurrence frequency of “keyword” (file number 5 document) is displayed at the top, and a document with a high appearance frequency of “kensaku” (file number 8 document) is displayed at the bottom. Is done. In this way, by using the versatility, it is possible to display a document that is obstructed by a generic word and buried in the lower level at the upper level.

  The merging unit 160 uses the Cost (d) obtained by the search unit 150 to determine the ranking of each document. The merging unit 160 may be configured by hardware (application circuit), or may be configured by an arithmetic device loaded with a program (software). The merge unit 160 sorts documents in descending order of Cost (d). At this time, the merging unit 160 collects predetermined information about each sorted document, the number of hits, and the like. The merge unit 160, for example, layout information (for example, image data) of a plurality of pages including pages including the file name, creation date, and search word of each sorted document, and layout information (for example, the first page) of the top page (first page). Image data), get the number of hits. At this time, it is preferable that the merging unit 160 obtains layout information (for example, image data) of a plurality of pages including a page where the appearance frequency of the search word is maximum.

  The merge unit 160 stores sort information (information about document arrangement order) and collected information (file names, etc.) in a predetermined storage area as necessary. Here, the predetermined storage area refers to an area to be accessed when the search result display unit 170 displays the search result on the display. Note that the merging unit 160 may directly pass the sort information and the collected information to the search result display unit 170 as necessary.

  The search result display unit 170 displays each document on the screen according to the ranking determined by the merge unit 160. The search result display unit 170 may be configured by hardware (application circuit) or may be configured by an arithmetic device loaded with a program (software). The search result display unit 170 first acquires information (sort information or the like) stored in a predetermined storage area, for example. Note that the search result display unit 170 may acquire sort information or the like directly from the merge unit 160 instead of collecting information (sort information or the like) stored in a predetermined storage area. The search result display unit 170 is composed of, for example, a web browser. Next, the search result display unit 170 displays each document on the screen based on the acquired information.

  7, 8, and 9 show examples of the layout of the search results displayed on the screen by the search result display unit 170. FIG. For example, a search window 171 and a search button 172 are arranged at the top of the screen, and buttons (file 173, page 174, thumbnail 175) for selecting a display form are arranged on the left side of the screen. Further, in the center of the screen, a file name, a creation date, and layout information (for example, image data) of one or a plurality of pages are arranged based on the sort information.

  For example, as illustrated in FIG. 7, the search result display unit 170 arranges layout information (for example, image data) of a plurality of pages including pages including the search word in a horizontal row for each document. As described above, when displaying the search result, not only the file name and the creation date but also the page layout is displayed, so that the user can search for a document while viewing the page layout. Furthermore, by displaying the layout of a plurality of pages all at once on the screen, the user can overlook the layout of the plurality of pages at a time, so that non-character information (for example, a figure or a table) around a sentence including a search term can be viewed. , A formula, a photograph, etc.), and a desired document can be found out. For example, in the case of “I want to search for a sentence containing a figure on the third page”, the user inputs a word that is predicted to be written around the figure as a search word, thereby obtaining a desired document. It is also possible to find out.

  In addition, for example, as illustrated in FIG. 8, the search result display unit 170 displays page layout information (for example, image data) including a search word one page at a time for each document. At this time, it is preferable that the search result display unit 170 displays page layout information (for example, image data) that maximizes the appearance frequency of the search word for each document page by page. Even when the page layout is displayed one page at a time for each document, the user can search for the document while viewing the page layout. Therefore, although the number of pages that can be viewed at a time is smaller than in the above case, a desired document can be obtained by using non-character information (for example, a figure, a table, a formula, a photograph, etc.) around the sentence including the search word as a clue. It is possible to find out.

  Further, for example, as shown in FIG. 9, the search result display unit 170 displays the layout information (for example, image data) of the top page of the document including the search word one page at a time for each document. In this case, the user can look over the layout of the top page of a very large number of documents at a time, so that the user can search for a document while viewing a large number of top pages.

  By the way, the search result display unit 170 does not include the page including the search word and the search word when simultaneously displaying the layout of the page including the search word and the layout of the page not including the search word on the screen. The page may be visually distinguished. For example, as illustrated in FIG. 7, the search result display unit 170 may highlight the edge 176 of the page including the search word. Further, the search result display unit 170 may selectively take out the document displayed as the search result and store it separately. For example, as shown in FIGS. 7, 8, and 9, when a retrieval icon 177 is displayed beside each document displayed as a search result, and the icon 177 is selected by the user, The document corresponding to the icon 177 may be saved separately.

[effect]
Next, the effect of the document search system 100 of this embodiment will be described.

  In the present embodiment, the versatility of each search word is calculated using the index 130 in which the appearance frequency for each word in units of n characters obtained from the document group to be searched is registered for each page. As described above, in this embodiment, since the versatility of each search term is derived, ranking display considering the versatility is possible. In addition, since the general degree of each search word is derived by calculation, the user does not need to worry about whether or not the word input as the search condition is a general word. In the present embodiment, using the index 130, the appearance frequency of each search word is calculated for each page that is a format break smaller than the document. As a result, it is possible to prevent a document in which search terms are not related to each other from being ranked high by simply making an exhaustive description. Accordingly, it is possible to reduce the influence of the general-purpose words and display a document including sentences in which the search terms are related to each other.

  In the present embodiment, since the search is performed in units of pages, the file size and the width of the description width do not affect the superiority of the document. Furthermore, when a plurality of search terms are used and all the search terms are distributed within one page, the search terms can be compared with each other even if the position of the search terms in the document is not grasped. Are likely to be related. Therefore, by performing a search in units of pages, it is possible to obtain a result equivalent to a search that considers the position of the search term in the document. Further, when searching in page units, position information in the document of the search word is not necessary in the first place, and therefore it is not necessary to execute a grep type search at the time of searching. Therefore, high-speed search can be performed.

  Further, in the present embodiment, in the example of FIG. 7, layout information (for example, image data) of a plurality of pages including a page including a search word is arranged in a horizontal row for each document. Thereby, the user can search for a document while looking at the layout of the page. Furthermore, by displaying the layout of a plurality of pages at a time, the user can look over the layout of the plurality of pages at once, so that non-character information (for example, a diagram, table, formula, It is also possible to search for a desired document using a photograph or the like as a clue.

<2. Modification>
[First Modification]
In the above embodiment, the division unit (n character unit) in n character division is two characters, but it may be one character or three or more characters. However, if the n character unit becomes too large, the n character unit may be the same as the number of characters in the search word or may be larger than the number of characters in the search word. It is preferably equal to or smaller than the statistical average value. For example, since the statistical average value of the number of characters in Japanese is 2.3 characters, when Japanese is used as a search word, the n character unit is preferably 2 characters or 3 characters. Furthermore, for example, in order to further improve the Japanese search accuracy, the n character unit should include not only two and three characters close to the statistical average value of the number of Japanese characters but also one character. Is preferred. Further, for example, since the statistical average value of the number of English characters is 5 characters, when English is used as a search word, the n character unit is preferably 5 characters.

[Second Modification]
Moreover, in the said embodiment and the 1st modification, although the unit (n character unit) of the division | segmentation in n character division | segmentation became one type, multiple types may be sufficient. FIG. 10 shows an example of the document search system 100 when there are a plurality of types of n character units. For example, as shown in FIG. 10, there are three types of n character units: k1 character units (k1 ≧ 1), k2 character units (k2> k1), and k3 character units (k3> k2). Good.

  At this time, the index registration unit 120 needs to perform the procedure from page division S104 to index merging S107 for each character type. For example, in the example of FIG. 11, the index registration unit 120 performs the procedure from page division S104 to index merging S107 for each k1 character unit, k2 character unit, and k3 character unit. Furthermore, the index registration unit 120 needs to register an index for each character type. For example, in the example of FIG. 11, the index registration unit 120 registers an index for each k1 character unit, k2 character unit, and k3 character unit. Therefore, the index 130 exists for each character unit. For example, in the example of FIG. 11, the index 130 exists for each k1 character unit, k2 character unit, and k3 character unit.

  Further, the search unit 150 needs to perform the procedure from the n character division S202 to the document weighting S205 for each character type. For example, in the example of FIG. 12, the search unit 150 performs the procedure from the n character division S202 to the document weighting S205 for each k1, character, and k3 character. The merging unit 160 selects the most appropriate weight from the document weights for each character.

  The n character unit preferably includes a character unit close to a statistical average value of the number of characters in the language of the search term that can be input. For example, when Japanese and English are used as search terms, the n character unit is 2 and 3 characters close to the statistical average value of the number of Japanese characters, and 5 close to the statistical average value of the number of English characters. It is preferable that it contains a character. Furthermore, for example, when it is desired to further improve the Japanese search accuracy, the n character unit preferably includes one character, two characters, three characters, and five characters.

[Third Modification]
In the above embodiment and its modifications, the unit of format separation when managing documents and indexes is a page. However, the present invention is not limited to this, and is, for example, a paragraph, chapter, or section. May be. Pages, paragraphs, chapters, and sections are groups of specific contents and serve as formal document structure markers. Therefore, by separating the document by pages, paragraphs, chapters, or sections, the document can be separated by semantic content.

  When paragraphs, chapters, or sections are used as document format separators, “page” may be read as “paragraphs, chapters, or sections” in the above-described embodiment and its modifications. For example, in the above-described embodiment and its modifications, the index registration unit 120 divides each acquired document in units of paragraphs, chapters, or sections, and acquires the paragraph, chapter, or section of each acquired document. N characters may be divided for each clause (steps S104 and S105). Further, for example, in the above-described embodiment and its modifications, the index registration unit 120 may register the word obtained by dividing the sentence in the divided index for each paragraph, chapter, or section of each document. Good. In this case, a word and the number of appearances are registered in the division index as a pair for each paragraph, chapter, or section of each document.

[Fourth Modification]
In the above-described embodiment and its modification, the case where the index registration unit 120 and the search unit 150 are configured by an arithmetic device loaded with a program is exemplified. In this case, the document search system 100 A mechanism for loading a program into the arithmetic device is provided. For example, the document search system 100 may include a reader that reads a program from a readable recording medium on which a program describing the contents executed by the index registration unit 120 and the search unit 150 is recorded. For example, the document search system 100 may include a communication system that acquires the above-described program via a network.

<3. Application example>
Hereinafter, application examples of the document search system 100 described in the above embodiment and its modifications will be described. The document search system 100 can be applied to a single document search apparatus 200 as shown in FIG. Further, as shown in FIG. 14, the document search system 100 can be applied to a system that inputs search conditions from the terminal device 300 to the document search device 200 via the external network 400. Further, as shown in FIG. 15, the present invention is applied to a system that uses a document search device 200 connected to the external network 400 to search a document group in the document storage device 500 connected to the external network 400. It is also possible. Further, as shown in FIG. 16, it is also possible to apply to a system in which a document group in the document storage device 500 connected to the LAN 600 is searched using the document search device 200 connected to the LAN 600. is there.

  A document search apparatus 200 illustrated in FIG. 13 corresponds to a function in which the function of the document search system 100 is realized by a single terminal device. For example, as illustrated in FIG. 13, the document search device 200 includes a control unit 210 that controls the entire document search device 200, a storage unit 220 that can store data used by the control unit 210, and input of search conditions. Input unit 230 and display unit 240 for displaying the search results. The control unit 210, the storage unit 220, the input unit 230, and the display unit 240 are connected to a common bus 250, for example. The storage unit 220 stores, for example, a document search program 221, a document storage unit 222, and an index 223 as shown in FIG.

  The document search program 221 causes the computer to execute a series of procedures executed by the index registration unit 120, the search unit 150, the merge unit 160, and the search result display unit 170. The document storage unit 222 corresponds to an aspect of the document storage unit 110. The index 223 corresponds to one aspect of the index 130. The control unit 210 loaded with the document search program 221 corresponds to an aspect of the index registration unit 120, the search unit 150, the merge unit 160, and the search result display unit 170.

  In the search system illustrated in FIG. 14, a terminal device 300 and a document search device 200 are connected via an external network 400. The document search apparatus 200 in FIG. 14 corresponds to the document search apparatus 200 in FIG. 13 in which the display unit 240 is omitted and a communication unit 260 is provided instead of the input unit 230. The communication unit 260 is a device that enables the document search device 200 to communicate with the terminal device 300 via the external network 400.

  The terminal apparatus 300 is an apparatus that receives input of search conditions, passes the received search conditions to the document search apparatus 200 via the external network 400, and presents search results to the user. For example, as illustrated in FIG. 14, the terminal device 300 receives a control unit 310 that controls the entire terminal device 300, a storage unit 320 that can store data used by the control unit 310, and input of search conditions. An input unit 330, a display unit 340 that displays search results, and a communication unit 350 that communicates with the document search apparatus 200 via the external network 400 are provided. The control unit 310, the storage unit 320, the input unit 330, the display unit 340, and the communication unit 350 are connected to a common bus 360, for example. For example, although not shown, the storage unit 220 stores software (for example, a web browser) that causes a computer to execute a series of procedures executed by the search result display unit 170. The document search program 221 in FIG. 14 causes a computer to execute a series of procedures executed by the index registration unit 120, the search unit 150, and the merge unit 160 described above. The document storage unit 222 corresponds to an aspect of the document storage unit 110. The index 223 corresponds to one aspect of the index 130. The control unit 210 loaded with the document search program 221 corresponds to an aspect of the index registration unit 120, the search unit 150, and the merge unit 160 described above. A control unit 310 loaded with software that causes a computer to execute a series of procedures executed in the search result display unit 170 corresponds to one mode of the search result display unit 170.

  The search system illustrated in FIG. 15 is a system in which a document search device 200 and a document storage device 500 are connected via an external network 400. The document search apparatus 200 in FIG. 15 corresponds to the document search apparatus 200 in FIG. 13 in which the document storage unit 222 is omitted and a communication unit 260 is further provided. The communication unit 260 is a device that enables the document search device 200 to communicate with the document storage device 500 via the external network 400.

  The document storage device 500 is a storage device accessed via a network. For example, as shown in FIG. 15, the document storage device 500 includes a control unit 510 that controls the entire document storage device 500, a storage unit 520 that can store data used by the control unit 510, and an external network 400. And a communication unit 350 that communicates with the document search apparatus 200 via the network.

  The search system illustrated in FIG. 16 is a system in which a document search device 200 and a document storage device 500 are connected via a LAN 600. The document search apparatus 200 in FIG. 16 corresponds to the document search apparatus 200 in FIG. 15 in which a communication unit 270 is provided instead of the communication unit 260. The communication unit 270 is a device that enables the document search device 200 to communicate with the document storage device 500 via the LAN 600.

  The document storage device 500 in FIG. 16 corresponds to the document storage device 500 in FIG. 15 in which a communication unit 550 is provided instead of the communication unit 530. The communication unit 550 is a device that enables the document storage device 500 to communicate with the document search device 200 via the LAN 600.

  As described above, the document search system 100 can be applied to various types of search systems.

  DESCRIPTION OF SYMBOLS 100 ... Document search system 110 ... Document storage part 120 ... Index registration part 121 ... Index structure 121A ... Word 121B ... File number 121C ... Page number 121D ... Appearance frequency 130 ... Index 140 ... Search condition Input unit 150... Search unit 160 160 Merge unit 170 Search result display unit 171 Search window 172 Search button 173 File 174 Page 175 Thumbnail 176 Edge 177 Icon 200 ... Document search device, 210, 310, 510 ... Control part, 220, 320, 520 ... Storage part, 221 ... Document search program, 222 ... Document storage part, 223 ... Index, 230, 330 ... Input part, 240, 340 ... Display unit, 250, 360, 540 ... Bus, 260, 270, 3 0,530,550 ... communication unit, 300 ... terminal apparatus, 500 ... document store, 600 ... LAN.

Claims (8)

A dividing unit that analyzes a given search condition and divides each search word included in the search condition in units of n characters (n ≧ 1);
The frequency of appearance for each word obtained by dividing each document in the search target document group by n characters is obtained by dividing each search word by using an index registered for each document format break. An extraction unit that extracts the appearance frequency for each word that is registered for each format break registered in the index;
While using the appearance frequency for each word extracted by the extraction unit, and not using the position information within the format break configured in the document of each search word, A document search system comprising: a weighting unit that calculates a generality of each search word and calculates a weight of each document using the appearance frequency and the generality obtained by the calculation. The document search system according to claim 1, wherein the format break is a page, a paragraph, a chapter, or a section. The document search system according to claim 1, further comprising a merge unit that determines a ranking of each document using the weight obtained by the weighting unit. The document search system according to claim 3, further comprising a search result display unit that displays each document according to the ranking determined by the merge unit. 5. The document search system according to claim 4, wherein the search result display unit displays a continuous layout of a plurality of format delimiters including a format delimiter having the highest appearance frequency in each document. The document search system according to claim 4, wherein the search result display unit displays a format-separated layout that maximizes the appearance frequency in each document. The n character unit includes a plurality of character units,
The index includes an index for each character unit included in the n character unit,
The dividing unit divides each search term in units of characters,
The extraction unit uses the index to extract the appearance frequency for each word obtained by the division in the division unit for each format break registered in the index and for each character unit,
The weighting unit uses the appearance frequency for each word obtained by the extraction in the extraction unit to determine the appearance frequency for each format term and each character unit, and the generality of each search word. The document search system according to any one of claims 1 to 6, wherein the weight of each document is calculated using the appearance frequency and the versatility obtained by the calculation. Analyzing a given search condition and dividing each search word included in the search condition in units of n characters (n ≧ 1);
The frequency of appearance for each word obtained by dividing each document in the search target document group by n characters is obtained by dividing each search word by using an index registered for each document format break. A second step of extracting the appearance frequency for each word obtained for each format break registered in the index;
While using the appearance frequency for each word extracted by the extraction unit, and not using the position information within the format break configured in the document of each search word, And a third step of calculating a general degree of each search word and calculating a weight of each document using the appearance frequency and the general degree obtained by the calculation.

Images