JPH07262199A - Retrieval result sorter - Google Patents

Abstract

PURPOSE:To improve classification efficiency by automatically selecting an appropriate classification axis by finding the evaluation value of the item value of an item sampled from a retrieval result and providing by classifying the retrieval result to a retriever based on a selected classification axis. CONSTITUTION:This sorter is equipped with an item/item value sampling means 3 which samples the item 31 and the item value 32 from the retrieval result 1, a classification axis selecting means 4 which selects the classification axis from sampled item 31 and item value 32, and a classification execution means 5 which classifies the retrieval result 1 based on the selected classification axis.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is a search result classification device for classifying search results, wherein an information search device selects a classification axis suitable for a search result and searches results along the classification axis. The present invention relates to a search result classification device that classifies automatically.

[0002]

2. Description of the Related Art In an information retrieval system, there are many results for one input, not only one result. When a large amount of electronic documents have been distributed and large-scale databases have become widespread as in recent years, the number of search results when conducting information searches on these databases has become large.

In the most basic information retrieval device, the number of retrieval results is displayed when a retrieval is performed, and the number of retrieval results is reduced to some extent by a logical operation or the like, and then the retrieval person sets the primary information to 1
After looking at one case, they are looking for the documents they need.

However, in such a retrieval device, it is only possible to determine whether or not the contents of the hit result set are appropriate by looking at each primary information, so it takes time to finally find the necessary information. There is a drawback that.

Therefore, there is also an apparatus which, when the number of results is obtained, classifies and displays the result set by an appropriate classification axis. Items such as "date" and "creator name" in the primary information can be used as the classification axis. By looking at such classified results, the searcher can reduce the amount of primary information that he or she needs to see in advance, which leads to efficient search.

[0006] In a conventional search device with a classification mechanism, a searcher himself or herself selects an appropriate classification axis, and the search device performs classification display using this selected classification axis.

[0007]

In the conventional search device for classifying and displaying the search results based on the classification axis, the searcher himself / herself needs to select the classification axis and is not familiar with the search mechanism or the search target data. There is a problem that the operation is difficult for general searchers.

[0008] In general, a general searcher does not recognize that displaying the results in a classified manner leads to an increase in the efficiency of the search, and even if he / she recognizes it, he / she does not know how to operate the classified display. In many cases, there is no knowledge about what kind of classification items there are. Even with such a search device having a result classification function, there is a problem that the classification function is not effectively utilized.

The present invention solves these problems.
The evaluation value of the item value extracted from the search result is obtained, and the appropriate classification axis is automatically selected. Based on this selected classification axis, the search results are classified and presented to the searcher to improve search efficiency. The purpose is to let.

[0010]

FIG. 1 shows a block diagram of the principle of the present invention. In FIG. 1, a search result 1 is a search result.

The item / item value extraction means 3 extracts the item 31 and the item value 32 from the search result 1. The classification axis selection means 4 uses the extracted item 31 and item value 3
The classification axis is selected based on 2.

The classification executing means 5 classifies the search result 1 based on the selected classification axis and outputs the classification result 7. The labeling means 6 extracts a keyword from the document having the item value 32 of the classification result 7 and adds it to the classification result 7.

[0013]

In the present invention, as shown in FIG. 1, the item / item value extracting means 3 extracts the item 31 and the item value 32 from the search result 1, and the classification axis selecting means 4 extracts the item 31 and the item value. A classification axis is selected from 32, and the classification execution means 5 classifies the search results 1 based on the selected classification axis.

At this time, a predetermined number is selected as the classification axis from the smallest of the two or more kinds of the item value 32 of the item 31 extracted by the classification axis selection means 4. Further, the item value 3 which is the classification axis selected by the classification executing means 5
When there is a meaningful order relationship in 2, the classification is performed according to the meaningful order relationship.

Further, when the item value 32 which is the selected classification axis does not have a meaningful order relation, the classification executing means 5 classifies the item value 32 by the appearance frequency of the item value 32. Further, the label adding means 6 extracts the keyword from the document having the classified result item value 32 and adds it to the classified result.

Further, the label adding means 6 extracts the keywords from the document having the classified result item value 32, arranges them in order of frequency, and adds them to the classified result. In addition, after the keywords are extracted from the document having the item value 32 of the classified result by the label adding means 6 and arranged in the order of frequency, the one having the highest frequency of use is positioned at the head and added to the classified result. I am trying.

Therefore, the item 31 extracted from the retrieval result 1
The evaluation value of the item value 32 of is determined and an appropriate classification axis is automatically selected, and the search result 1 is classified based on the selected classification axis, or the keyword of the document having the classification axis item value 32 is added. By doing so, it is possible to classify and display the search result 1 according to the automatically selected classification axis, display it by adding a keyword, and present it to the searcher to improve the search efficiency. .

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Next, the construction and operation of an embodiment of the present invention will be described in detail with reference to FIGS.

FIG. 2 shows a flowchart for explaining the operation of the present invention. This is a procedure for automatically selecting the classification axis in the configuration of FIG. 1 and outputting the classification result 7. In FIG. 2, S1 receives the search result. For example, FIG. 3 is received as the search result 1 obtained by searching a database (not shown) with the keyword “translation”. Search result 1
Is a list of item values 32 of items 31 as shown in FIG.

In step S2, items and item values are extracted from the search results. In S3, one item is taken out. This retrieves one item 3 from the search result 1 in FIG. 3, for example, the item “author name”.

In S4, the types of item values corresponding to the items are counted. This is the item value 32 of the item 31 extracted in S3.
, The type of item value 32 of the item 31 “author” in FIG. 3, for example, “Taro Suzuki”, “Jiro Yamada”, ...

In step S5, it is determined whether the item is the last item. In the case of YES, since the number of item values 32 of all items 31 has been counted, the process proceeds to S6. On the other hand, in the case of NO, S3 and S4 are repeated for the next item 3, and the number of item values 32 is counted.

Search result 1 is obtained by the above S1 to S5.
The number of types of item values 32 for each item 31 is counted, and the number of types of item values is obtained as shown in the right column of FIG. 4, for example.

At S6, an item is selected. In S7, the classification is performed with the item selected in S6 as the classification axis. These S6,
In S7, a predetermined number of items 31 are selected from the items of less than 2 in the number of items 32 (the number of item value types in FIG. 4) of each item 31 counted in S4, and the selected items 31 are selected. Search result 1 is classified as a classification axis. For example, as shown in FIG. 4 which will be described later, the number of item value classifications in FIG. 4 is two or more and the smallest number of items is three items 31. Classification code: 2 (number of item value types) Author name: 5 Publication year: 9 For each of the three selected classification axes (classification code, author name, publication year), the items are classified in descending order of appearance frequency (evaluation value) of the item value of the search result 1 as shown in the classification result 7 of FIG. Here, the reason why the item having two or more items and having a small number of items is selected is that it is difficult for the searcher to determine whether or not the item having a large item value 32 is effective for narrowing down, and the item is excluded from the display area. Is.

In step S8, the classification result is output to the output device.
As described above, the item value 3 of the item 31 of the search result 1 in FIG.
When the two types are counted, as shown in FIG. 4, the number of item value types is obtained for each item 31, and when the number of types of the item value 32 is 2 or more and a small number is selected as a predetermined number, 3 is selected. As shown, “field code”, “author name”, and “issue year” are obtained as the item 31, and the classification result 7 is output as shown in the drawing by arranging the item values 32 in the order of appearance frequency.
As a result, the classification axis is automatically selected from the search result 1, the classification result 7 classified according to the classification axis can be displayed, and can be automatically presented to the searcher, and the search efficiency can be improved.

FIG. 3 shows an example of the search result of the present invention. The search result 1 shows a part of the result obtained by searching a database (not shown) based on the keyword "translation". here,
Item 31 is the author name, title, year of publication, source, field code, text, etc. in the upper row. The items described below these items 31 are the item values 32. For example, the item values 32 “Taro Suzuki”, “Jiro Yamada” of the item 31 “Author name” are ...

FIG. 4 shows an example of sorting search results according to the present invention for each item. This means that the item values 32 are arranged in the horizontal direction for each item 31 of the search result 1 in FIG.
The number of types (number of item value types) is counted (see S4 in FIG. 2).

FIG. 5 shows an example of the classification result of the present invention. This is because the number of item value types in the arrangement example of FIG. 4 is 2 or more and a small number of items are selected, here, three items are selected and the item value 32 of each item 31 and the evaluation value of the item value 32, here, It is a compilation of appearance frequencies. That is, the number of item value types in FIG. 4 is 2 or more, which is small.-Item 31 with item value type 2 "Field code" -Item 31 with item value type 5 "Author" -Item 31 with item value type 9 “Issuance year” is selected in order (selected as a classification axis), and the appearance frequency of the item value 32 of each item 31 is tabulated as an evaluation value as illustrated.

Next, according to the order shown in the flow chart of FIG. 6, the structure and operation for labeling will be described in detail with reference to FIGS. 7 to 9 based on the classification result 7 of FIG.

In FIG. 6, S11 determines whether or not the item value of the classification result has a code. This determines whether there is a code of the item value 31 of the classification result, for example, there is a code in the classification result 7 of FIG. 5, here a field code. Yes
In this case, the processing of adding the keyword after S12 as a label is performed. On the other hand, in the case of NO, the process ends.

In step S12, the body of the document having the item value is extracted from the search result. This retrieves, for example, the body of the document of search result 1 in FIG. 3 having the corresponding item value 32. In step S13, keywords are extracted from the text extracted in step S12. For example, a character string having a high appearance frequency is extracted as a keyword.

In step S14, the keyword is added as a label to the classification result 7. For example, as shown in FIG.
As a label for the item value 32, the keyword "processor" or the like extracted from the illustrated text is added.

In step S15, the classification result with the label is output. For example, the label of FIG. 7 is added to the classification result 7 of FIG. 5 and output. As described above, when the item value 32 of the classification result 7 of FIG. 5 described above has a code (for example, a field code), a keyword (for example, a character string having a high appearance frequency) is extracted from the text of the document having the item value 32 of the code. ) Is extracted, added to the classification result 7 as a label, and output.
This allows the searcher to display the classification result 7 in FIG. 5 and the keyword extracted from the text in association with the item value 3 and improve the search efficiency.

FIG. 7 shows an example of labeling item values according to the present invention. This is a keyword extracted from the text as a label in association with the code of the item value on the classification axis. Here, three are selected in descending order of appearance frequency.

FIG. 8 shows item / item value learning data of the present invention. This is stored in a file for each database, and the usage frequency is stored in association with the classification axis. Here, the classification axis "author name" is used 25 times, and the classification axis "field code" is used 18 times.

FIG. 9 shows an example of the classification result of the present invention. This reflects the learning data of FIG. 8 on the classification result 7 of FIG. 5 (the search result 7 in which the number of item value types is 2 or more and the items are selected in ascending order). That is, according to the learning data held for each database in FIG. 7, the usage frequency of the author name is 25
Since the field code is used more frequently than 18 times, the field code and the author name in the search result 7 in FIG. 5 are replaced with each other.

Next, according to the order shown in the flow chart of FIG. 10, the classification result 7 shown in FIG. 5 is labeled with high-use keywords as shown in FIG. 12 based on the learning keywords shown in FIG. The operation will be described in detail.

In FIG. 10, a step S21 decides whether or not there is a code in the item value of the classification result. This determines whether or not there is a code in the item value 31 of the classification result, for example, there is a code in the classification result 7 in FIG. 5, here a field code. YE
In the case of S, the process of adding the keyword after S22 as a label is performed. On the other hand, in the case of NO, the process ends.

In step S22, the body of the document having the item value is extracted from the search result. This retrieves, for example, the body of the document of search result 1 in FIG. 3 having the corresponding item value 32. In step S23, keywords are extracted from the text extracted in step S22. For example, a character string having a high appearance frequency is extracted as a keyword.

In S24, if there is a learning keyword among the keywords in the text, this is brought to the beginning of the order. For example, if the learning keywords shown in FIG. 11 provided for each database have a frequency of 10 or more (communication, database), and the learning keyword is found in the keyword list extracted in S23, the category axis item is given first. Come on. For example, regarding the search result 7 of FIG. 5, even if the learning keyword is registered in the corresponding database as shown in FIG. 11 and the evaluation value is low in the keyword list, the learning keyword is as shown in FIG. To the beginning of the label (for example, “communication” and “database” extracted from the learning keyword are added to the beginning as shown in FIG.

FIG. 11 shows an example of learning keywords of the present invention. The learning keywords are provided for each database and arranged in the order of appearance frequency of the keywords extracted from the text (see FIG. 5). Then, the learning keywords of FIG. 11 having a frequency of 10 or more and the appearance of the keywords of FIG. If it exists even if the frequency is low, the learning keyword is added first. Each time the learning keyword is used, the frequency is incremented by one.

FIG. 12 shows an example of labeling reflecting the high-frequency keywords of the present invention. In this case, the item value 32 in which the frequency of the learning keyword in FIG. 11 is, for example, 10 or more is taken out and added to the head of the label in FIG.

[0043]

As described above, according to the present invention,
The evaluation value of the item value 32 of the item 31 extracted from the search result 1 is obtained, and an appropriate classification axis is automatically selected, and the search result 1 is classified based on the selected classification axis, or the item value of the classification axis is further selected. Since the structure of adding the keyword of the document having 32 is adopted, the search result is classified according to the automatically selected classification axis and displayed, or further added with the keyword and displayed. To improve the search efficiency.

[Brief description of drawings]

FIG. 1 is a principle block diagram of the present invention.

FIG. 2 is a flowchart explaining the operation of the present invention.

FIG. 3 is an example of a search result of the present invention.

FIG. 4 is an example of sorting search results according to items of the present invention.

FIG. 5 is an example of a classification result of the present invention.

FIG. 6 is a labeled flowchart (No. 1) of the present invention.

FIG. 7 is an example of labeling of item values according to the present invention.

FIG. 8 is learning data of items / item values of the present invention.

FIG. 9 is an example of a classification result of the present invention.

FIG. 10 is a flowchart with a label according to the present invention (Part 2).
Is.

FIG. 11 is an example of a learning keyword of the present invention.

FIG. 12 is an example of labeling in which high-frequency keywords of the present invention are reflected.

[Explanation of symbols]

 1: Search result 2: Search result classification device 3: Item / item value extraction means 31: Item 32: Item value 4: Classification axis selection means 5: Classification execution means 6: Labeling means 7: Classification result

Claims (7)

[Claims] 1. A search result classification device for classifying search results, wherein items (31) and item values (32) from search result (1).
An item / item value extraction means (3) for extracting a classification axis, a classification axis selection means (4) for selecting a classification axis from the extracted item (31) and item value (32), and the selected classification axis. A search result classification device, characterized by further comprising a classification execution means (5) for classifying the search results (1). 2. The classification axis selecting means (4) selects a predetermined number as a classification axis from the smallest of two or more kinds of the item value (32) of the extracted item (31). The search result classification device according to claim 1. 3. The classification executing means (5), when there is a meaningful order relationship in the item value (32) which is the selected classification axis, performs classification according to the meaningful order relationship. The search result classification device according to claim 1 or 2. 4. The classification executing means (5) classifies the item value (32) according to the frequency of appearance when the item value (32) which is the selected classification axis does not have a meaningful order relationship. The search result classification device according to claim 1 or 2, characterized in that. 5. A label adding means (6) for extracting a keyword from a document having an item value (32) of the classified result (7) and adding it to the classified result (7) is provided. The search result classification device according to any one of claims 1 to 4, which is characterized. 6. A label adding means (6) for extracting keywords from a document having an item value (32) of the classified result (7), arranging them in order of frequency and adding them to the classified result (7). Claims 1 to 4 characterized in that
Search result classification device described in. 7. After extracting keywords from a document having an item value (32) of the classified result (7) and arranging them in order of frequency, the previously used one having a high frequency of use is positioned at the head and the classification is performed. 5. A label adding means (6) for adding to the result (7), comprising: a label adding means (6).
Search result classification device described in.