JP2010277568A - Retrieval device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a retrieval device which obtains the retrieval result that an operator narrows down a plurality of documents to the appropriate number of documents. <P>SOLUTION: A document retrieval repetition means 14 of the retrieval device 10 sequentially retrieves documents for retrieval from the lowest ratio to the highest ratio of retrieval word hit ratio set by a retrieval word hit ratio setting means 13 or the opposite direction of ratio. When the number of documents for retrieval becomes equal to or less than the matching number specified by a matching number specification means 11 in the case of the direction from the lowest ratio to the highest ratio, or when the number of documents for retrieval exceeds the matching number in the case of the opposite direction of ratio, retrieval processing stops, and a document retrieval result output means 15 outputs a retrieval result. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a search device that searches for a desired document from various documents such as academic documents, magazine articles, newspaper articles, patent documents, and various contents of a homepage.

  Document retrieval is routinely performed in various places. Search devices for this purpose range from specialized computers to those that execute simple application software on a CPU (Central Processing Unit) such as a personal computer. Such search devices can be broadly classified according to search engines, search methods, and search targets.

  As search engines, various engines such as an index type engine, an RDB (relational database) type engine, a full-text search engine, and a concept search engine have been proposed. As a search method, a Boolean operation method that performs logical operations such as AND, OR, NOT, and NEAR, a correlation coefficient order that is a vector in a vocabulary space, and a percentage of the number of hit words among a plurality of specified search words A ranking type method of displaying in order is known. As a search target, there is known a target for the entire document, a target for a page, or a target for a character string in a predetermined range such as a document or a sentence that is a constituent unit of a page. .

  By the way, when a large number of documents are mechanically searched by a search device, documents having different contents that are not assumed at the stage of selecting a search word as a search keyword are often shown as search results. This often depends on the method of selecting a search word, but it is also caused by the fact that one document may not necessarily have the contents according to one theme. For example, in a diary-like document existing on the Internet, miscellaneous contents about social events and life that happened in one day are described in a considerable length of text, and finally it finally reaches the main subject. Such a document has a phenomenon in which the content intended by the search operator does not exist, or even if it exists, the content is relatively high as a search result regardless of the sparse content. Therefore, conventionally, it has been necessary to check the contents of a relatively large amount of documents extracted corresponding to a search word one by one over a huge amount of time.

In order to solve such a problem, as a first related technique of the present invention, a search device that extracts a search word from a search sentence specified by a searcher and searches for a matching document from the related document by a ranking type technique. Has been proposed (see, for example, Patent Document 1). In the first related technique, the contents of the search result are examined, and the document requested by the searcher is selected as the relevant document. Then, a conforming document designating unit (not shown) designates a plurality of conforming documents from the search result according to the selection. Similarly, a related word extraction unit (not shown) extracts a related word from the relevant document designated by the searcher and inputs it to the related document search unit (not shown). The related document search unit searches for documents including related words from the document group as related documents in a ranking type, and displays the search results. As a result, the matching document designated by the searcher appears in the upper search.

  On the other hand, regarding the problem that noise is generated in the search result due to miscellaneous contents being described in the document, the second problem is that the search target is not a whole document but a sentence unit. Related techniques have been proposed (see, for example, Patent Document 2). In the second related technique, a document file stored in a database is divided into sentence units at the time of retrieval. Then, an AND search is performed with a plurality of search terms on all document sentences.

JP 2006-251935 A (paragraph 0039, paragraph 0040, FIG. 2) JP 2001-195416 A (paragraphs 0017 and 0018, FIG. 1 and FIG. 2)

  Of these, in the first related technique, when the number of conforming documents obtained as a search result is enormous, the searcher needs to carefully examine and select the required number of documents from among them. The problem of need remains. In addition, the search operator needs to specify the search term several times in the process of narrowing down the number of conforming documents to a desired number, or to execute a search for related documents a plurality of times.

  On the other hand, the second related technique has a problem that the processing time of the search device is considerably long. That is, in the second related technique, the search device searches all sentences in all documents one by one. As an example, let's assume a gazette concerning a patent application as a search target. Suppose that there is a database that contains 100,000 documents as documents. Assuming that the number of sentences per patent gazette is 100, 10 million sentences are searched for in the entire database. When searching for such an enormous amount of text, it takes a long time to search, and it is difficult to realize the search by an information processing apparatus such as a general personal computer or server. In addition, if such a search is put to practical use, a large amount of cost is required.

  SUMMARY OF THE INVENTION An object of the present invention is to provide a search device that can obtain a search result narrowed down to the number of documents that the search operator makes appropriate from among a plurality of documents.

  Another object of the present invention is to provide a search device that can more reliably obtain a search result intended by a search operator.

  In the present invention, (b) a matching number designating unit for designating a matching number as a number of documents to be finally searched among a plurality of documents to be searched, and (b) the final searching object described above. A search word designating means for designating a predetermined number of search words as terms to be included in the document, and (c) the highest from the lowest ratio among the search word hit ratios set by the search word hit ratio setting means When the documents to be searched are sequentially searched in the direction of percentage or from the highest percentage to the lowest percentage, and the number of obtained documents is in the direction from the lowest percentage to the highest percentage. When the number of conforming cases specified by the above-mentioned conforming number specifying means is less than or equal to the number of conforming cases specified above, or when the number of conforming cases is exceeded in the direction from the highest ratio to the lowest ratio, Document search repeating means for stopping the processing, and (d) the document at the stage where the search processing is stopped by the document search repeating means in the direction from the lowest ratio to the highest ratio, or the highest ratio to the lowest ratio The search apparatus includes a document search result output unit that outputs a document at the immediately preceding stage that exceeds the number of matching cases in the direction of.

  As described above, according to the present invention, a document to be searched in the direction from the lowest ratio to the highest ratio among the search word hit ratios set by the search word hit ratio setting means or vice versa. Search sequentially. The number of documents obtained when the number of documents obtained in the direction from the lowest percentage to the highest percentage is less than or equal to the number of documents specified by the number-of-conforms specifying means, or vice versa. The search process is stopped when it exceeds. Therefore, regardless of the number of documents to be searched, it is possible to select an appropriate number of documents according to the search purpose without requiring human intervention. This eliminates the need for a screening operation as a visual confirmation operation by the search operator. As a result, the search operator can concentrate on examining the selected document if necessary.

It is a claim corresponding | compatible figure of the search device of this invention. It is a block diagram showing the outline | summary of the structure of the search device by embodiment of this invention. It is a flowchart showing the flow of the search process by the search device of this Embodiment. It is a top view showing the principal part of the window displayed on the display which comprises a search word and the number-of-conformity designation | designated part in this Embodiment. It is the block diagram which actualized the sentence hit rate process part by this Embodiment. It is the flowchart which showed the flow of the process of the hit ratio process part in a sentence in this Embodiment. It is a top view showing the principal part of the window displayed on the display as one form of the output of a search result in this Embodiment. It is the top view which showed an example of the window displayed on the display, when confirming the sentence which the search operator hit in this Embodiment.

  FIG. 1 is a diagram corresponding to claims of the search device of the present invention. The search device 10 of the present invention includes a matching number specifying means 11, a search word specifying means 12, a search word hit ratio setting means 13, a document search repetition means 14, and a document search result output means 15. Here, the matching number designation means 11 designates the matching number as the number of documents to be finally searched among a plurality of documents to be searched. The search term designating unit 12 designates a predetermined number of search terms as terms to be included in the final document to be searched. The search word hit ratio setting means 13 sets a plurality of search word hit ratios as a ratio of the number of search words specified by the search word specifying means 12 included in each document to be searched. The document search repetition unit 14 is configured to search from the search rate hit ratio set by the search word hit rate setting unit from the lowest rate to the highest rate, or from the highest rate to the lowest rate. The documents that become are sequentially searched. Then, when the number of obtained documents is equal to or less than the above-mentioned number of matching specified by the matching number specifying means 11 when the direction is from the lowest ratio to the highest ratio, or from the highest ratio described above, When the number of matching cases is exceeded when the direction is low, the search process is stopped. The document search result output means 15 is a document at a stage where the search processing is stopped by the document search repeat means 14 when the direction is from the lowest ratio to the highest ratio, or when the direction is from the highest ratio to the lowest ratio. The document at the previous stage that exceeds the number of matching cases is output as a search result.

  <Embodiment of the Invention>

  Next, an embodiment of the present invention will be described.

  FIG. 2 shows an outline of the configuration of the search device according to the embodiment of the present invention. The search device 100 according to the present embodiment includes a main control unit 101 that controls the entire device. The main control unit 101 includes a CPU (Central Processing Unit) 102 and a memory 103 that stores a control program for controlling the apparatus. The memory 103 may be a semiconductor memory such as a ROM (Read Only Memory), or may be a storage medium such as a magnetic disk, an optical disk, or a DVD (Digital Versatile Disk). The main control unit 101 is connected to the following units and performs these controls. However, at least a part of each unit described below may be realized by software by the CPU 102 executing a control program stored in the memory 103.

  The document storage unit 104 stores a search target document. However, the search target document does not need to be stored in the document storage unit 104. The search target document may be stored in a recording medium such as a CD (Compact Disc) or a DVD, or may be stored in a server on a network connected via a communication unit (not shown).

  The search term / applicable number designation unit 105 is a device that designates or inputs a search term for searching a document stored in the document storage unit 104 and designates the number of matches. For example, the search word / matching number designation unit 105 can be configured using a display such as a keyboard or a liquid crystal display (not shown). The search term / matching number designation unit 105 can input not only the search terms but also the equivalent terms as terms for the search. An equivalent word is a term having a meaning equivalent to that of a search word, and may be a similar word having a similar meaning.

  The number of matching cases refers to the maximum number of documents to be finally searched among a plurality of documents to be searched. The search apparatus 100 according to the present embodiment performs a first-stage search process for selecting a document to be searched in detail from documents to be searched, and for each character string constituting the document selected in this process. A second-stage search process for checking the hit rate of the search term is performed. The number of matching cases is the maximum number of documents obtained by the first-stage search process, and a default value is adopted or the search operator specifically specifies it.

  The search engine 106 performs a search process. The search engine 106 may be an arbitrary search engine such as an index type engine, an RDB (relational database) type engine, a full-text search engine, or a concept search engine.

  The search repeater 107 repeats the search process using the search engine 106 until a predetermined condition is satisfied. The search repetition unit 107 includes an intra-document hit rate addition unit 108, a search purpose relevant number determination unit 109, and a relevant number comparison unit 110.

  The in-sentence hit rate processing unit 111 searches the document stored in the document storage unit 104, and performs a process of calculating the hit rate for each sentence in the document that falls within the number of matching cases. Here, the sentence means each character string when the document is divided according to a predetermined rule from one end to the other end of the document. Typically, each sentence separated by punctuation points may be a sentence, or a paragraph in which these sentences are gathered may be a single sentence. Further, an arbitrary number of adjacent sentences may be collected and a character string constituting them may be used as one sentence. Which range of character strings is used as one sentence differs depending on the document creation environment such as language type, language usage, and document creator. In this embodiment, as an example, a case will be described in which each sentence separated by punctuation is a sentence.

The answer output unit 112 is a device that outputs a final search result and the result of the first-stage search process. The answer output unit 112 may be a display such as a liquid crystal display, or may be one that prints out on paper such as a printer. Of course, the search results can be output by voice, or output to the outside using a communication means (not shown)
It may be stored in a storage medium. In the present embodiment, a case where the answer output unit 112 is configured with a display will be described. In this case, the display constituting the search word / applicable number designation unit 105 and the component parts of the apparatus can be used together.

  FIG. 3 shows the flow of search processing by the search device. Here, a search for a patent publication will be taken as an example and described with reference to FIG.

  First, the search device 100 enters a state in which a search operator can set search conditions from the search word / matching number designation unit 105 as a preparation stage for executing a search (step S201). In this state in which the search condition can be set, the search operator can specify, for example, the search target range in the document stored in the document storage unit 104, the number of matching cases, and the search term. When the operation for instructing the search execution is performed (step S202: Y), the main control unit 101 sets an initial value of the hit rate in the document based on the search condition set in step S201 (step S203). A search by the search engine 106 is executed (step S204).

  FIG. 4 shows a main part of a window displayed on the display constituting the search word / applicable number designation unit. A window 121 opened on a display (not shown) is provided with a text input field 122, and below it, a search word designation field 123 and a similar word designation field 124 are provided.

  When the search term is set in step S201 in FIG. 3, the input area 125 of the search term designation field 123 is designated. In FIG. 4, the designated input area 125 is indicated by a mesh. When a search word is input to the text input field 122 using an input means such as a keyboard (not shown) and the arrow button 126 is pressed, the search word moves to the input area 125. FIG. 4 shows a state where the term “paper” is input to the input area 125 as an example of a search term.

  For example, it is assumed that the search operator inputs the term “paper” as a search term into the search term designation column 123 and inputs the term “plain paper” as a similar term to the similar term designation column 124. In this case, after inputting the term “paper” in the first line of the search word designation field 123, the “similar word 1” in the similar word designation field 124 is designated as the input area 125, and the text input field The term “plain paper” is entered in 122. Then, the arrow button 126 may be pressed. Similarly, similar words of “paper” such as “sheet” and “cut paper” can be input at a location where a similar word such as “similar word 2” is designated in the similar word designation field 124.

  Although the case where the search term “paper” and its similar words are input has been described above, a plurality of search terms are usually set. It is not always necessary to input similar terms for each search term.

  On the right side of the arrow button 126 in the window 121, a hit rate setting column 127 in the publication (in the literature) is provided. The in-document hit rate (hereinafter abbreviated as hit rate) may be set automatically, or the stepwise hit rate can be set manually by the search operator. In the example shown in FIG. 4, automatic setting is selected.

  For example, it is assumed that the hit rate is automatically set in a state where search terms are input as three terms A, B, and C. In this case, if none of the three search terms A, B, and C exists in the corresponding document, the hit rate is 0%. If any one of the three search terms A, B, and C exists (hits) in the document, the hit rate is 33%. If two of the three search terms A, B, and C are present in the document, the hit rate is 66%. If all three terms A, B, and C are present in the document, the hit rate is 100 percent. However, in this example, the hit rate is rounded down.

  Similarly, assume that four terms A, B, C, and D are set as search terms and the hit rate is automatically set. In this case, there are four hit rates: 0%, 25%, 50%, 75%, and 100%. When a similar word is set as a search word, even if the search word does not hit, if any of the corresponding similar words hits, the hit rate is calculated as a search word hit. Of course, when “sheet” and “cut paper” are set as similar words in the search term “paper”, even if all of these hit in one document, the hit rate of the search term “paper” is high. It is not a premium.

  When the search operator manually sets the hit rate, as in the case of automatic setting, a numerical value such as “0, 33, 66, 100” or a plurality of arbitrarily increasing numerical values is set in the hit rate setting field 127. To enter. At this time, if the maximum numerical value is set to “100”, it is possible to obtain a search result including all of the search terms specified in the search term specification column 123 (logical product). Of the hit rate, 0 percent has no meaning in search. Accordingly, the low level hit rate as the lowest hit rate is set to a value such as 1% in accordance with the search purpose. Depending on the search device 100, the low level hit rate during manual operation may be set to a small fixed value such as 1%. In addition, the hit rate may be set to be increased in small increments up to 100% in increments of 5%, for example, regardless of the number of search terms.

  Returning to FIG. 3, the description will be continued. In the initial setting of the hit rate in step S203, the lowest value other than the value “0” among the set hit rates is set as the initial value. First, all documents are sequentially searched one by one with the set search word and similar word, and it is determined whether the number of documents that are equal to or higher than the initial value of the hit rate is equal to or less than the number of matching cases (step S205). This determination process is performed by the matching number comparison unit 110. For example, assuming that there are 10,000 documents to be searched in total, the search operator sets the number of matching cases to 10 or less. Assume that three terms A, B, and C are set as search terms, and there are a total of 500 documents including any one or more of these three terms A, B, and C. In this example, when the hit rate as an initial value is 33%, it is determined that the number of target documents (the number of hit documents) is not less than the number of matching cases (step S205: N).

  Of course, in order to perform the determination in step S205, it is not necessary to search all documents to be searched for the three search words A, B, and C (including similar words). When searching for a document to be searched, in this example, when the number of matching cases exceeds 10, the search processing for the document is stopped, and the determination in step S205 may be performed.

  In the case of the example just described, it has been found that the hit rate is not less than the number of matching cases for 33% (step S205: N). Therefore, in this case, it is determined whether or not the hit rate can be further added (step S206). In this example, the hit rate can be added to 66% and 100% (Y). Therefore, the hit rate is increased by one step to 66% (step S207). Thereafter, the process returns to step S204, and the search by the search engine 106 is executed again.

  In this example, a document including two or more of the three search terms A, B, and C is searched this time. Then, it is determined whether or not the number of target documents obtained as a result is equal to or less than the number of matching documents (step S205).

  In this example, it is assumed that even when the hit rate is 66%, the number of corresponding documents exceeds the matching number of 10 (step S205: N). Then, the determination in step S206 is performed, and the hit rate is increased to 100% in step S207. As a result, it is assumed that three documents that include all three search words A, B, and C among all the documents to be searched are included. In this case, the condition that the number of applicable documents is 10 or less, which is the number of conformance, is satisfied (step S205: Y). Therefore, the search apparatus 100 ends the first-stage search process for selecting a document to be searched. Then, for each character string constituting the document selected by the above processing, the in-sentence hit rate processing is executed as the second-stage search processing for checking the hit rate of the search word (step S208).

  On the other hand, even when the hit rate is raised to 100%, if the number of applicable documents exceeds the matching number of 10 (step S205: N), it is determined whether the hit rate can be added (step S205: N). Step S206). In this case, the hit rate cannot be further increased (N). Therefore, an error display indicating that the process cannot be shifted to the second-stage search process is performed (step S209), and the process ends (END). At this time, a display such as “Add a search term or reset the search condition” may be output to the display constituting the answer output unit 112. In this case, the search operator performs the process of step S201 again while changing the search condition, and then executes the search again.

  FIG. 5 shows an embodiment of the intra-sentence hit rate processing unit. The in-sentence hit rate processing unit 111 includes a sentence set creation unit 131, an in-sentence search unit 132, and a storage unit 133. Here, the sentence set creation unit 131 performs a process of creating a set of sentences for each of the documents that have been answered in the first-stage search process (hereinafter referred to as answer documents) with the number of matching cases or less. This is a process of dividing one document into character strings of a predetermined unit. Here, as an example, each sentence separated by punctuation is processed as a sentence. Therefore, specifically, a process of dividing the document by the phrase “.” Is performed.

  The in-sentence search unit 132 searches each sentence created by the sentence set creation unit 131 using the search terms designated by the search term / matching number designation unit 105 in FIG. The hit rate for the search word for each sentence is stored in the storage unit 133.

  FIG. 6 shows the flow of processing in the sentence hit rate processing section. This will be described with reference to FIGS.

  The sentence set creation unit 131 of the intra-sentence hit rate processing unit 111 first sets the parameter K indicating the number of the response document to “1” (step S221), and the Kth response document (in this case) A set of sentences in the first response document) is created (step S222). For example, if the first answer document is composed of 100 sentences, it is defined as a set of first to 100th sentences.

Next, the parameter S representing the sentence number is set to “1” (step S223). Then, the hit rate for the search word set by the search word / matching number designation unit 105 in the S-th sentence (in this case, the first sentence) is calculated (step S224). The calculation of the hit rate is the same except that what is performed for each document in the first-stage search processing is performed for each sentence. For example, A, B, to the three search terms and C, and A ₁ of the similar word (even word) for search terms A, and A ₂ is set. In this example, for example, if a sentence does not include any of these three search terms and similar terms, the hit rate is 0 percent. If any of the search word A, the similar word A _1, or the similar word A ₂ is included and the search words B and C are not included, the hit rate is 33%. If the search terms A, B, and C are included in one sentence, the hit rate is 100%.

As described above, when the in-sentence search unit 132 calculates the hit rate of the S-th sentence, it is stored in the storage area associated with the K-th answer document in the storage unit 133 in association with the sentence (step S225). ). Then, if the parameter S at this time has not reached the sentences of the S _MAX of the last sentence of the answer document of the K (step S226: N), adds the parameter S by "1" (step S227). Thereafter, the process returns to step S224, and this time, the hit rate related to the second sentence is calculated.

In the same manner, the hit rate is calculated in order from the first sentence in the first answer document, and these results are stored in the storage unit 133 associated with the first answer document. When the sentence processing for the last (maximum) number is completed (step S226: Y), the Kth answer document (in this case, the first answer document) that has been processed is the last (maximum) number. It is determined whether or not the answer document is (step S228). If the answer document of the last (maximum) number is not the answer document (K _MAX answer document) (N), the parameter K is incremented by “1” (step S229). Then, returning to step S222, in this example, the hit rate for the second answer document is calculated.

In the same manner, hit rates are calculated in order from the first sentence in the second answer document, and these results are stored in the storage area associated with the second answer document in the storage unit 133. . In this way, when the processing of the K _MAX answer document which is the last (maximum) number is completed (step S228: Y), the hit rates of the sentences in the storage area of each answer document are rearranged in descending order. (Step S230), and the process ends (END). Instead of performing such rearrangement, when reading from the storage unit 133, reading processing may be performed in descending order of hit rate.

  Returning to FIG. 3, the description will be continued. When the processing in the sentence hit rate processing unit 111 described with reference to FIG. 6 is performed (step S208), since the search is normally completed, the answer is output (step S210), and the processing is terminated (END).

  FIG. 7 shows a main part of a window displayed on a display as one form of output of search results. In the display window 151, an answer column 152 indicating search results is arranged, and a search history column 153 is arranged therebelow.

In this example, since the number of matching cases is 3 as a result of the search, those having the highest hit rate in the sentence (hereinafter referred to as optimum sentence) are displayed in the answer column 152 for these three cases. According to the answer column 152, it is understood that Japanese Patent Application Laid-Open No. 08-314305 filed by Company F has the highest hit rate when viewed in sentence units. This indicates that all of the search terms A, B, and C were included in the specific sentence of this publication (document). However, for the search word A, either of the similar words (equivalent words) A ₁ and A ₂ may be included.

The second highest hit rate is a sentence included in the publications (documents) of Company C and Company S. These mean that two of the search words A, B, and C are included in a specific sentence. However, for the search word A, either of the similar words (equivalent words) A ₁ and A ₂ may be included.

  In the search history column 153, numbers representing the history of the search and the terms used in the search are displayed corresponding to these numbers.

  FIG. 8 shows an example of a window displayed on the display when checking a sentence hit by the search operator. The window 161 shown in FIG. 8 can be expanded by clicking one of the publications (documents) displayed in the answer field 152 as shown in FIG. However, the contents of the window 161 shown in FIG. 8 are not corresponding to the example shown in FIG.

  In the window 161, a sentence selection field 162 is arranged at the top, and below that, a hit sentence field 163 and an example display field 164 showing the full text of the part shown as an example in the gazette (document). Is arranged. Below the example display field 164, a search button 165 and a search word input field 166 are arranged. When a search phrase is entered in the search phrase input field 166 and the search button 165 is pressed, a desired phrase can be searched from the text information shown in the example display field 164.

  By the way, the sentence selection field 162 is provided with a “return to optimal sentence” button 167 and buttons 168 and 169 for advancing or returning the optimum rank in the same publication (document). The search operator can change the contents of the sentence displayed in the hit sentence column 163 by appropriately operating these buttons 167 to 169.

  Further, a visual device 171 for distinguishing from the remaining character strings is provided for the character strings corresponding to the search words and their similar words (equivalent words) in the sentences displayed in the hit sentence column 163. Here, the case where the visual device 171 is underlined is shown, but the corresponding character string itself is colored in a color such as red, or the font is different from the remaining character strings, or is displayed in reverse video. May be.

  According to the search device 100 of the present embodiment as described above, there are the following advantages.

  The search device 100 automatically repeats the search from the low hit rate to the high hit rate, and presents documents having a preset number of conformances or less to the search operator. Recent search engines are high-speed, and for example, there are many known search engines that can ignore a decrease in practical performance even if the search is repeated about 20 times. Therefore, the search operator can acquire a document group in which the search target is narrowed down to some extent as a search result without significantly reducing the search speed.

  In the case of the present embodiment, since the search method is a ranking type and the search target is a sentence, even if there is a miscellaneous description in the document, the contents described locally can be accurately extracted. it can. In addition, since a device used for the search can use a general device such as a general personal computer or a server, an inexpensive device configuration can be realized.

  In particular, in the case of the search work of patent documents shown as an example in the embodiment, there are various purposes, and there is a standard number that the search result is sufficient if there are as many answers as there are according to the purpose. Exists. Taking the search work based on the patent invalid material search as an example, if the final result is about 10 useful materials, then the purpose of the search is achieved. In the present embodiment, such a standard number of cases can be set as the number of matching cases, and the increase in search time due to this is almost negligible. Therefore, in addition to the ranking type search process with a hit rate of what percentage, it is possible to greatly reduce the load required for the search operator process while maintaining the reliability of the search process.

  In addition, in the present embodiment, each type of document obtained based on the number of matching cases is classified into sentences and a ranking type search is performed. For this reason, it is possible to easily find the most notable sentence. In addition, by performing a search divided into sentences, it is possible to obtain a search result that is not affected by the description content of a place where noise may occur in the content such as a location where the prior art is described. .

  In the present embodiment, the search for patent documents has been described by way of example. However, the search device of the present invention can be applied to various documents such as academic documents, magazine articles, newspaper articles, and various contents on a homepage. Naturally, it can be done. In addition, when dividing a document into a plurality of character strings for each group, it is needless to say that the document is not limited to a unit using the punctuation points shown in the embodiment.

  Further, as shown in the example display column 164 of FIG. 8, when a document to be searched is divided into sentences, it is necessary to add a punctuation at a line feed position or to replace a specific identification symbol with a punctuation. Pre-processing may be performed.

  Furthermore, in the present embodiment, documents to be searched are sequentially searched in the direction from the lowest ratio to the highest ratio among the search word hit ratios set by the search word hit ratio setting means. The search process is stopped when the number of matching cases specified by the above-mentioned matching number designation means becomes less than the number of matching cases, and the document at that stage is output as a search result. In addition to this, the documents to be searched are sequentially searched in the direction from the highest ratio of the search word hit ratio set by the search word hit ratio setting means to the lowest ratio, and the number of obtained documents is When the number of matching cases is exceeded, it is possible to stop the search process and output a document at the previous stage that exceeds the number of matching cases as a search result.

DESCRIPTION OF SYMBOLS 10,100 Search apparatus 11 Matching number designation | designated means 12 Search term designation means 13 Search term hit ratio setting means 14 Document search repetition means 15 Document search result output means 101 Main control part 102 CPU
DESCRIPTION OF SYMBOLS 103 Memory 104 Document storage part 105 Search term / applicable number designation part 106 Search engine 107 Search repetition part 110 Conformity number comparison part 111 In-sentence hit rate processing part 112 Reply output part 121, 151 Window 131 Sentence set creation part 132 Search in sentence Section 133 Storage section

Claims (6)

A matching number specifying means for specifying a matching number as the number of documents to be finally searched among a plurality of documents to be searched;
A search word designating means for designating a predetermined number of search words as terms to be included in the final search target document;
A search word hit ratio setting means for setting a plurality of search word hit ratios as a ratio of the number of search words specified by the search word specifying means included in each document to be searched;
The documents to be searched are sequentially searched from the lowest rate to the highest rate in the search term hit rate set by the search word hit rate setting means, or from the highest rate to the lowest rate. , When the number of obtained documents is less than or equal to the number of conformances designated by the conformity number designation means when the direction is from the lowest percentage to the highest percentage, or the direction from the highest percentage to the lowest percentage A document search repetition means for stopping the search process when the number of matching cases is exceeded at the time of
The document at the stage where the search process is stopped by the document search repeating means when the direction is from the lowest rate to the highest rate, or the number of matching cases is exceeded when the direction is from the highest rate to the lowest rate. A search apparatus comprising: a document search result output means for outputting a document at a previous stage as a search result. Character string dividing means for dividing a document designated by the document search result output means into a plurality of character strings for each group from one end to the other end of the document,
A character string hit ratio calculating means for calculating a ratio of a predetermined number of the search words specified by the search word specifying means for each character string divided by the character string dividing means;
The character string hit ratio calculation means further comprises a character string search result display means for displaying the search result of the search word with respect to the character string of the corresponding document from the one containing the predetermined number of the search words. The search device according to claim 1, wherein:   3. The search apparatus according to claim 2, wherein the character string segmenting unit classifies each sentence divided by punctuation marks in the document as the character string.   3. The search apparatus according to claim 2, wherein the character string segmenting means classifies a predetermined number of sentences separated by punctuation marks in the document as the character string.   A search for instructing re-input of a predetermined number of search terms designated by the search term designation means when it is determined that the number of documents including all of the search terms designated by the search term designation means exceeds the number of matching cases The retrieval apparatus according to claim 1, further comprising word re-input instruction means.   2. The search device according to claim 1, wherein an arbitrary number of equivalent words of the term can be designated as a search term to be designated by the search term designation means.

Images