JP3678615B2 - Document search apparatus and document search method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a document search apparatus and a document search method for searching for a desired document in accordance with a search condition, and in particular, the degree to which each document matches the search condition and a bibliographic item accompanying each document, for example, a date for a newspaper article. The search results can be sorted and displayed according to combinations such as the new order.
[0002]
[Prior art]
2. Description of the Related Art In recent years, a document ranking method has been drawing attention in which the degree of matching between a document and a search condition is obtained based on the appearance frequency of a search word in the document and the results are rearranged and displayed in descending order. In addition, bibliographic items attached to documents, for example, newspaper articles, specify dates as sorting conditions and display them with priority over articles with newer dates. Articles with the same date are ordered in descending order of suitability with search conditions. Flexible search such as display has been realized.
[0003]
As shown in FIG. 13, the conventional document search apparatus extracts word frequency information of words included in a dictionary 1302 from document data 1301 of newspaper articles to be searched, and stores word frequency information extraction means for storing in a word frequency index 1304 1303, bibliographic item information such as date / paper name code is extracted from document data 1301, bibliographic item extracting means 1305 is stored in the bibliographic item index 1306, and the user inputs a search request character string including a search condition and a sort condition Search request input means 1307 for search, word frequency calculation means 1308 for finding the appearance frequency in the document of the search word included in the search condition by examining the word frequency index 1304, and logical operation means for performing logical operation between record sets 1309, a fitness calculation means 1310 for calculating the fitness between the search condition and each record, a sort information acquisition means 1311 for acquiring bibliographic information for sorting specified in the sort condition, Result sorting means 1312 for sorting the record list of search results according to bibliographic information and fitness, and result display means 1313 for displaying the search results are provided.
[0004]
The word frequency index 1304 stores the appearance frequency of dictionary words in the search target document by the extraction operation of the word frequency information extraction unit 1303.
[0005]
FIG. 14 is a flowchart showing a search processing procedure in a conventional document search apparatus. The document data 1301 is composed of a plurality of records (documents) delimited by a record delimiter, and each record is composed of a plurality of fields delimited by a field delimiter. FIG. 3 shows a specific example of the document data. In the newspaper article data, the field delimiter is “^ F”, the record delimiter is “^ R”, the paper name code, the date, and the article body. is there.
[0006]
The word frequency information extraction unit 1303 scans the document data 1301 in advance, counts how many times the word registered in the dictionary 1302 appears in the article body field of each record, and records in which the word appears The number and the total number of records are stored in the word frequency index 1304.
[0007]
The bibliographic item extracting unit 1305 scans the document data 1301 in advance and stores the contents of the bibliographic item field of each record in the bibliographic item index 1306.
[0008]
First,
Step 1401: The user inputs a search request character string using the search request input means 1307. The search request character string consists of two parts, a search condition and a sort condition. FIG. 15 shows a specific example of a search request character string, where “Matsushita AND New Product” is a search condition, and an article that includes both search terms “Matsushita” and “New Product” in the article text. "@HIDUKE @SHIMEI" means that the search results are arranged in the newest date order, and if the same date, the paper name codes are arranged in ascending order. If the date and paper name codes are the same, arrange them in order of fitness.
[0009]
Step 1402: The word frequency calculation means 1308 applies to all search terms,
Step 1403: Referring to the word frequency index 1304, for the search word included in the search condition input by the search request input means 1307, the number of records in which the word appears in the article body, the internal number of each record, the relevant number in each record The word appearance frequency and the total number of records are calculated.
[0010]
Step 1404: The logical operation means 1309 performs a logical operation between the record sets output by the word frequency calculation means 1308.
[0011]
Step 1405: The goodness-of-fit calculation means 1310 targets all search result records,
Step 1406: For each record output by the logical operation means 1309, the degree of matching (Rel) with the search condition is calculated by, for example, (Equation 1).
Rel = Σ (TFi · IDFi) (Σ is added for i)
IDFi = 1-log ₂ (DFi / ND) (Equation 1)
However, TFi represents the appearance frequency of the search word Wi in the record, DFi represents the number of records in which the word Wi appears, and ND represents the total number of records.
[0012]
Note that the method of calculating the fitness is not limited to (Equation 1).
[0013]
Step 1407: The sort information acquisition means 1311 refers to the bibliographic item index 1306, and sorts the values of the bibliographic items corresponding to the sort condition input from the search request input means 1307 of each record output by the fitness calculation means 1310. Obtain as information.
[0014]
FIG. 6 shows an example of the output contents of the sort information acquisition means 1311. The date and the page code value are acquired as sort information.
[0015]
Step 1408: The result sorting means 1312 sorts and outputs the output of the sort information obtaining means 1311 using a plurality of bibliographic items obtained as sort information as a sort key. At this time, if there are records having the same value of all the bibliographic items, the records are rearranged in descending order of fitness.
[0016]
FIG. 16 is an example of the output contents of the result rearranging means 1312.
[0017]
Step 1409: The result display means 1313 shapes the output of the result rearranging means 1312 and presents it to the user.
[0018]
[Problems to be solved by the invention]
However, in the conventional configuration, the value of the bibliographic item specified in the sort condition is given priority over the relevance as the sorting key, so the document with the lower relevance is higher and the document with the higher relevance is higher. There is a problem that it may be ranked lower, and a desired document cannot be searched efficiently.
[0019]
For example, this corresponds to the document ranked at the bottom in FIG. 8 (record internal number 10).
[0020]
The present invention solves such problems of the prior art, and allows the user to limit the range of the fitness level while emphasizing the value of the bibliographic item specified in the sort condition as a sort key. the document does not enter the specified goodness of fit range, more by rank lower, providing a document retrieval system which can find efficiently a desired document, also aims to provide a document search method It is said.
[0021]
[Means for Solving the Problems]
Therefore, in the document search apparatus of the present invention, as a search request character string, in addition to the search condition and the sort condition, search request input means for inputting a fitness range specification, and the document suitability are within the designated fitness range. Intermediate result classifying means for assigning different classification flags depending on whether or not it is applicable is provided.
[0025]
In the document search method of the present invention, the search condition is satisfied from the stored document data in response to a search request that specifies the search condition, the sort condition, and the fitness range indicating the degree of matching with the search condition. Search the document, calculate the fitness of each detected document, obtain the sort information of each document to sort each document according to the sorting condition, and specify the fitness of each document specified in the search request Compared with the range, the classification flag indicating whether it falls within the range is assigned to each document. Each document is first sorted by the classification flag, and if the value of the classification flag is the same, the sort information If the sort information is the same, the sort information is sorted and displayed in order of suitability.
[0026]
Therefore, the document outside the scope of matching level specified by the user, more can be ranked lower, of specifying a sort criteria, a low fitness document-level, high adaptability document to the lower The problem of being ranked can be avoided.
[0027]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[0028]
(First embodiment)
FIG. 1 is a block diagram showing a configuration of a document search apparatus according to the first embodiment of the present invention.
[0029]
Similar to the conventional apparatus (FIG. 13), this apparatus extracts word frequency information of words included in the dictionary 102 from the document data 101 of newspaper articles to be searched and extracts word frequency information stored in the word frequency index 104. Means 103, bibliographic item extracting means 105 for retrieving bibliographic information such as date / paper name code from the document data 101 and storing it in the bibliographic item index 106, search request input means 107, word frequency calculating means 108, logical operation means 109, A fitness level calculation unit 110, a sort information acquisition unit 112, a result sorting unit 114, and a result display unit 115 are provided, and the fitness level of each record calculated by the fitness level calculation unit 110 is converted into a relative value with respect to the maximum value. Relative fitness calculation means 111 that outputs to sort information acquisition means 112 and records whose fitness values do not fall within the specified fitness range from the search results output from sort information acquisition means 112 Intermediate results, excluding and a cutback means 113.
[0030]
The flowchart of FIG. 2 shows a search processing procedure in the first embodiment. The document data 101 is composed of a plurality of records (documents) delimited by record delimiters, and each record is composed of a plurality of fields delimited by field delimiters. FIG. 3 shows a specific example of document data, which is newspaper article data having a field delimiter “^ F”, a record delimiter “^ R”, and three fields of paper name code, date, and article body.
[0031]
The word frequency information extraction means 103 scans the document data 101 in advance, counts how many times the word registered in the dictionary 102 appears in the article body field of each record, and records in which the word appears The number and the total number of records are stored in the word frequency index 1304.
[0032]
Further, the bibliographic item extracting means 105 scans the document data 101 in advance and stores the contents of the bibliographic item field of each record in the bibliographic item index 106.
[0033]
First,
Step 201: The user inputs a search request character string using the search request input means 107. The search request character string consists of three parts: a search condition, a sort condition, and a fitness range specification. FIG. 4 shows a specific example of a search request character string, where “Matsushita AND New Product” is a search condition, and an article containing both search terms “Matsushita” and “New Product” in the article body. "@HIDUKE @SHIMEI" is a sort condition, and the search results are arranged in the newest date order, and if the same date, the paper name code is arranged in ascending order. "70:" part is a fitness range specification, which means that only articles having a relative fitness of 70 or more with respect to an article having the highest fitness are included in the result. If the date and paper name codes are the same, arrange them in order of fitness. It is also possible to specify both the lower limit and the upper limit of the fitness range specification, such as “$ 70: 90”, and to specify that the articles whose fitness is 70 or more and 90 or less are included in the result, or only the upper limit Is possible.
[0034]
Step 202: The word frequency calculation means 108 targets all search terms,
Step 203: Referring to the word frequency index 104, for the search word included in the search condition input by the search request input means 107, the number of records in which the word appears in the article body, the internal number of each record, the relevant number in each record The word appearance frequency and the total number of records are calculated.
[0035]
Step 204: The logical operation means 109 performs a logical operation between the record sets output by the word frequency calculation means 108. FIG. 5 shows an example of output contents of the logical operation means 109 in the case of the search request character string shown in FIG. 4, and a record set in which both “Matsushita” and “new product” appear is obtained.
[0036]
Step 205: The fitness level calculation means 110 calculates all the search result records as targets. Step 206: For each record output by the logical operation means 109, the fitness level with the search condition is calculated by, for example, (Equation 1). To do.
[0037]
Step 207: The relative fitness level calculation unit 111 converts the fitness level of each record output from the fitness level calculation unit 110 to a value obtained by dividing the record by the maximum value and multiplying by 100.
[0038]
Step 208: The sort information acquisition unit 112 refers to the bibliographic item index 106 according to the sort condition input by the search request input unit 107, and sorts the bibliographic item values of each record output by the relative fitness calculation unit 111. Obtain as information. FIG. 6 shows an example of the output contents of the sort information acquisition means 112, and the date and the page code value are acquired as sort information.
[0039]
Step 209: The intermediate result cut-off means 113 targets all the records output from the sort information acquisition means 112,
Step 210: Check whether the fitness of the record corresponds to the fitness range specification input from the search request input means 107,
Step 211: Exclude records that are not applicable.
[0040]
FIG. 7 is an example of contents output from the intermediate result cut-off means 113 when the fitness range designation is 70 or more.
[0041]
Step 212: The result sorting means 114 sorts the output of the intermediate result rounding means 113 using a plurality of bibliographic items acquired as sort information as a sort key, and conforms when all the bibliographic items have the same value. Sort and output in descending order. FIG. 8 shows an example of the output contents of the result rearranging means 114. Since the result documents are arranged in ascending order of the paper name code with the newest date and the articles whose fitness is outside the specified range are excluded, the user can efficiently find a desired document.
[0042]
Step 213: The result output means 115 shapes the output of the result rearranging means 114 and presents it to the user.
[0043]
As described above, in this document search apparatus, it is possible to display a document that does not fall within the fitness range from the searched documents, so that a desired document can be searched efficiently.
[0044]
In addition, when a search result document is cut off by relevance, it is possible to sort the search results once by relevance and cut off documents whose relevance does not reach the specified value. Since the number of documents as a result is large, the sort processing load for this document becomes extremely heavy. On the other hand, in the method of this embodiment, it is only necessary to check whether or not the document suitability falls within the specified suitability range for each document. Light processing. Therefore, the document search result can be displayed quickly.
[0045]
Note that the sort information acquisition in step 208 may be performed on the document after YES in step 209, that is, after the search result is cut off. The amount of work can be reduced.
[0046]
(Second Embodiment)
In the second embodiment, a document search apparatus that displays documents by distinguishing them according to the rank of fitness will be described.
[0047]
As shown in FIG. 9, this apparatus assigns different classification flags to the search result records output from the sort information acquisition unit 912 depending on whether the fitness value falls within the specified fitness range. Intermediate result classification means 913 is provided. Further, unlike the first embodiment, there is no intermediate result cut-off means. Other configurations are the same as those of the first embodiment (FIG. 1).
[0048]
FIG. 10 is a flowchart illustrating a search processing procedure according to the second embodiment. Here, the procedure up to step 1008 is the same processing procedure as in the first embodiment.
[0049]
Step 1009: The intermediate result classifying means 913 targets all records output from the sort information acquiring means 912,
Step 1010: Check whether the fitness of the record corresponds to the fitness range designation input from the search request input means 907,
Step 1011: For records that do not fall within the fitness range, “2” is assigned as the value of the classification flag,
Step 1012: “1” is assigned as the value of the classification flag for the record corresponding to the fitness range.
[0050]
FIG. 11 shows an example of the output contents of the intermediate result classifying means 913.
[0051]
If both the lower limit and the upper limit are specified as the fitness range, the intermediate result classification means 913 further subdivides records that do not fall within the fitness range, and records that exceed the upper limit as the value of the classification flag. It is also possible to give “3” as the value of the segmentation flag to records that are less than “2”.
[0052]
Step 1013: The result sorting means 914 sorts the output of the intermediate result sorting means 913 in descending order of the value of the classification flag. If the values of the classification flag are the same, a plurality of bibliographic items acquired as sort information Are sorted as sort keys, and if there are records with the same bibliographic item values, the records are sorted in descending order of suitability.
[0053]
FIG. 12 shows an example of the output contents of the result rearranging means 914. Articles with the newest date, the result documents arranged in ascending order of the paper name code, and articles whose fitness is outside the specified range are ranked lower than articles whose fitness is within the specified range. Can efficiently find a desired document.
[0054]
Step 1014: The result output means 915 shapes the output of the result rearranging means 914 and presents it to the user.
[0055]
As described above, in the document search apparatus according to this embodiment, all the searched documents can be displayed by being classified into those that fall within the fitness range and those that do not. Depending on the purpose of the search, the user can end the document search only by looking at the documents of the category that falls within the suitability range, or when one omission is not allowed, such as when searching for patent documents In addition, it is possible to examine each document that falls outside the fitness range.
[0056]
【The invention's effect】
As apparent from the above description, the document search apparatus and document retrieval process of the present invention, it is possible to fit it to the document outside the scope specified by the user, it ranks more lower.
[0057]
By doing so, it is possible to avoid the problem that a document with a low fitness level is ranked higher and a document with a higher fitness level lower than when a sort condition is specified, and a desired document can be searched efficiently. It becomes possible.
[0058]
In addition, by converting the fitness of each document into a relative value with respect to the maximum value and specifying the fitness level range in the search request as a relative value, an appropriate fitness level range can be easily specified.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of a document search apparatus according to a first embodiment of the present invention;
FIG. 2 is a flowchart showing a search processing procedure according to the first embodiment;
FIG. 3 is a diagram showing an example of document data;
FIG. 4 is a view showing an example of a search request character string in the first embodiment;
FIG. 5 is a diagram showing an example of output contents of the logical operation means in the first embodiment;
FIG. 6 is a diagram showing an example of output contents of sort information acquisition means in the first embodiment;
FIG. 7 is a view showing an example of output contents of an intermediate result cut-off means in the first embodiment;
FIG. 8 is a diagram showing an example of output contents of a result rearranging unit in the first embodiment;
FIG. 9 is a block diagram showing a configuration of a document search device according to the second embodiment of the present invention;
FIG. 10 is a flowchart showing a procedure of search processing according to the second embodiment;
FIG. 11 is a diagram showing an example of output contents of intermediate result classification means in the second embodiment;
FIG. 12 is a diagram showing an example of output contents of a result rearranging unit in the second embodiment;
FIG. 13 is a block diagram showing a configuration of a conventional document search apparatus;
FIG. 14 is a flowchart showing a conventional search processing procedure;
FIG. 15 is a diagram showing an example of a search request character string;
FIG. 16 is a diagram illustrating an example of output contents of a result rearranging unit.
[Explanation of symbols]
101, 901, 1301 Document data
102, 902, 1302 dictionary
103, 903, 1303 Word frequency information extraction means
104, 904, 1304 word frequency index
105, 905, 1305 Bibliographic item extraction means
106, 906, 1306 Bibliographic Index
107, 907, 1307 Search request input means
108, 908, 1308 Word frequency calculation means
109, 909, 1309 Logical operation means
110, 910, 1310 Conformity calculation means
111, 911 Relative fitness calculation means
112, 912, 1311 Sort information acquisition means
113 Intermediate result rounding off
114, 914, 1312 Result sorting means
115, 915, 1313 Result display means
913 Intermediate result classification means

Claims (6)

 In a document search device that searches stored document data according to a search condition and displays the search results sorted according to the sort condition, the search condition, the sort condition, and a fitness indicating the degree of matching with the search condition Search request input means for inputting a search request character string consisting of a range specification, search means for searching for a document satisfying the search conditions, and fitness calculation for calculating the fitness of each document searched by the search means Means for obtaining sort information for sorting each retrieved document according to the sort condition, and the degree of conformity for which the suitability of each retrieved document is specified in the range Search result classifying means for assigning to each document a classification flag indicating whether or not it falls within the goodness-of-fit range compared with the range; and the search result classification Each document with the classification flag output from the stage is first sorted by the classification flag. If the values of the classification flag are the same, the documents are sorted by the sort information, and the sort information is the same. A document search apparatus comprising: search result sorting means for rearranging in order of the fitness in case, and search result display means for displaying the search results sorted by the search result sorting means. The search means searches for documents that match the search conditions and calculates the appearance frequency of search words in each document, and the fitness calculation means is based on the search word appearance frequencies calculated by the search means. The document search apparatus according to claim 1 , wherein the fitness of each document is calculated. The search means searches for a document that matches the search condition, calculates the number of documents in which the search word appears, and the appearance frequency of the search word in each document, and the matching degree calculation means calculates the search word in each document. The document search apparatus according to claim 1 , wherein the fitness of each document is calculated based on the appearance frequency of the document and the number of documents in which the search word appears. The relevance calculation means includes an absolute relevance calculation means for calculating the relevance of each document, and a relevance value of each document calculated by the absolute relevance calculation means relative to the highest relevance among them. Relative conformity calculation means for converting into a document, the conformity calculation means outputs the relative adaptability represented by the relative value as the adaptability of each document, and the search request input means The document search apparatus according to claim 1 , wherein the range specification is performed based on the relative fitness.  In a document search method for searching stored document data according to a search condition and displaying the search result sorted according to the sort condition, the search condition, the sort condition, and a fitness indicating the degree of matching with the search condition In response to a search request designating a range, a document satisfying the search condition is searched from accumulated document data, the fitness of each detected document is calculated, and the documents are rearranged according to the sort condition. The sort information of each document for the purpose is acquired, and the degree of matching of each document is compared with the range of the degree of matching specified in the search request, and a classification flag indicating whether the document falls within the range is given to each document The documents are first sorted by the sorting flag. If the values of the sorting flag are the same, the sorting information is sorted and the sorting information is the same. Document search method characterized in that the display is rearranged in order of relevance to the case of Tsu. As the fitness of each detected document, a relative fitness for the highest fitness among the fitness of each document is calculated, and a range of fitness can be specified by the relative fitness in the search request. The document retrieval method according to claim 5 , wherein the document retrieval method is performed.

Images