JP2022079010A - Information selection support system - Google Patents

Abstract

To solve a problem in an information selection support system (so-called search system) including an Internet search engine in which the selection result is displayed as a list of display contents by text or image or match rate, in which a person has to determine whether the selection result is the desired one or not by checking all or part of the contents of the result list by, even when the survey analysis is performed by artificial intelligence, a person has to confirm the results resulting in an increase of work for a person causing one of the obstacles to adopting artificial intelligence.SOLUTION: It is not the selected text itself or part of it, but a sentence is created as the steps to reach a selection result as a basis to reach a reason or conclusion. By displaying or reading aloud the sentence or displaying images, the selection result can be determined by looking only at the reasons and grounds for selection instead of judging the content of the text by looking at all or part of it.SELECTED DRAWING: Figure 1

Description

The present invention is an information selection support system, by creating a reason for selection, a sentence as a basis, etc., from the reason, the basis, without reading, seeing, or listening to all or a part of all the sentences, etc. selected by a person. It is an effective technology applied to information selection support systems and information selection systems based on artificial intelligence, which make it easier to judge whether the selection is correct or not, and whether it is correct or not cannot be determined without looking at all the selection results. Regarding

In the conventional information selection support system and document selection system by artificial intelligence, sentences are selected from the high and low degree of matching of keywords converted using keywords and thesaurus, and the degree of matching and matching points are displayed. Then, the method of providing as a selection result is common.

The person who performs the information retrieval makes a visual judgment as to whether or not the result matches the purpose of the information retrieval, and in the information retrieval system using artificial intelligence, for efficiency improvement. There is a contradiction that more work is required to determine the outcome of the artificial intelligence adopted in.

Also, as a basis, the current information retrieval system and the information retrieval system using artificial intelligence are based on the degree of matching of keywords converted using keywords and thesaurus, so even if the number of matches is large, the order of appearance is There were cases where it was different or there were things that could not be judged only by matching, such as numbers, and it was not a reason for whether the result display content was correct.

In addition, as the basis for selection, there are some that used the background of manual investigation and selection result in the past, but if the conclusions are the same and the background is different, it is not the basis for selection.

Japanese Patent Application No. 2007-097973: Information selection support system, terminal device, information selection support device, information selection support method, program selection reason is a keyword matching condition (claim 1), or the keyword is a thesaurus dictionary database result matching condition ( Claim 3), and the reason for selection is the match or the number of matches of words as in the case of a general search engine or artificial intelligence. Even if the words match, the contents do not always match because the order is different, so it cannot be treated as a reason for selection. There is a need to. Japanese Patent Application No. 2005-168590: There is a description that it is based on an arbitrary sentence selected in the underlying document file and the reason for selection, but the reason for selection is any word or any word described in claim 5. When multiple keywords are selected, the function is to automatically guide the search formula for searching other literature files by combining those keywords, and it is based on the keyword selection combination as in the case of Japanese Patent Application No. 2007-097973. And Japanese Patent Application Laid-Open No. 2010-224984: The publication number of the citation is the past notice of reasons for refusal, and the reason for using it as a citation example is given. It can be the basis for selection only when it is similar or matches, but since it cannot be linked to the creation of the basis other than the prepared sentence like the notice created by a person, there is no basis sentence itself. , I can't show the rationale. Japanese Patent Application Laid-Open No. 2008-117010: Comparing similarities based on past patent decisions and rejection decisions, and drawing conclusions based on past cases when similar or matching ones can be extracted. Therefore, it is not possible to show the rationale in new investigations and searches that have no past cases.

In an information selection support system (so-called search system) such as an Internet search engine, the selection results are displayed in a list by text or image, a match rate, or the like. Whether or not these are the contents that you want to search can only be determined by a person checking all or part of the contents of the result list. With artificial intelligence, even if research and analysis are performed, it is necessary for humans to confirm the results, and even if artificial intelligence is adopted, it causes more work by humans, which is one of the obstacles to adopting artificial intelligence. Are

Further, in the conventional information retrieval system and the information retrieval system by artificial intelligence, the match degree or the number of matches of a keyword (mainly noun only) and a word converted by a synonym obtained from a thesaurus database is used as a basis. There is a high percentage of content that does not match just by matching keywords. This is because when it becomes a sentence, the meaning of the sentence changes depending on the order of the vocabulary.

Writing a procedure to reach a conclusion as a basis for judging whether the selection is correct without peer review of the result for judging whether the information acquisition result by a person is correct.

The outline of the invention of the present application will be described as follows. That is, the present invention uses text recognition by natural language processing for text input to be searched in an information selection system such as an artificial intelligence, an in-server search system, or an Internet search system, and then divides the text sentence into phrases. , Compare the phrases and judge whether they match or not from the relational system of each word. The sentence is created based on the procedure that the match is judged by whether or not all the key phrases of the input document are included, and it is judged from which phrase matches the input sentence. If the input text contains numbers and it is judged that the numbers need to be compared, it is judged whether the numbers are within the range of the comparison target, and whether the unit system related to the numbers matches. Confirm and create a sentence based on the procedure of calculation. In other words, create a sentence based on the procedure of determining whether or not it matches the input sentence content and determining whether or not the number is within the target range. By adding these grounds as the reason for selecting the information selection result, it is possible to judge the certainty of the selection result without looking at the content of the selection result itself.

In a general information selection system (so-called search system), it is not possible to know whether or not the selected information is desired unless the contents of each of the selected candidate lists are confirmed. Therefore, the very costly work of checking all the candidates in order until the result list hits the desired content is narrowed down to the required selection target without checking the content of the result list selected by the present invention in detail. Because it can be done, the cost of confirming the selection result can be reduced.

Overall view of creating, displaying, and reading aloud the text of the procedure that is the basis and reason for the information selection system Flow of creating sentences in the survey / selection results, the procedure for deriving the results as the reason / ground for selection Reason, example of creating evidence document

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. However, the present invention can be implemented in many different embodiments and should not be construed as limited to the description of this embodiment. In addition, the same element shall be assigned the same number throughout the embodiment.

In the following embodiments, the method or system will be mainly described, but as will be obvious to those skilled in the art, the present invention can also be implemented as a computer-usable program. Therefore, the present invention can take an embodiment as hardware, an embodiment as software, or an embodiment of a combination of software and hardware. The program can be recorded on any computer-readable medium such as a hard disk, CD-ROM, optical storage device or magnetic storage device, semiconductor storage device, and the like.

Further, in the following embodiments, a general computer system can be used. The computer system that can be used in the embodiment includes a central processing unit (CPU), a main storage device (main memory: RAM), a non-volatile storage device (ROM), a coprocessor, an image accelerator, a cache memory, and input / output control. It is equipped with hardware resources generally provided in a computer system such as a device (I / O). Further, an external storage device such as a hard disk device and a communication means capable of connecting to a network such as the Internet can be provided. The computer system includes various computers such as a server computer and a client computer, a personal computer, a workstation, and a mainframe computer, and various portable information terminals such as a tablet computer and a smartphone via a network.

FIG. 1 is a schematic diagram of a system for creating, displaying, and reading aloud the grounds and reasons in an information selection system and artificial intelligence having an information selection system function. The information selection reception block decomposes the received sentences (hereinafter referred to as input sentences) into words and parts of speech. Next, the text content is compared. Here, if the input sentence decomposed into words / parts of speech contains numbers or numbers and units, the numbers / units are compared at the same time as the sentence content comparison. Obtain the part of the matching sentence by comparing the sentence contents. In number / unit comparison, numbers and units are combined and compared to derive the relationship between input and target.
Create a procedure for investigating these matching points and relationships as a sentence of the reason for the selection decision, and display or read aloud as the basis / reason together with the result.

Generally, in the information selection system (so-called search system), the keyword to be searched is input and converted by the morphological data as the keyword, and it is judged whether or not it matches by the number of hit keywords, etc., or from the text. Words are extracted using CABOCHA (Japanese dependency analyzer) and MECAB (Japanese morphological analysis engine), and the same processing as information selection by the following keywords is performed. After all, information is selected based on the number of matching keywords. In this case, the reason for selection is the magnitude of the number of matches with the keyword (or the thesaurus-converted keyword). Actually, there is a high possibility that different arrangements of characters including particles, verbs, etc. will have completely different meanings, and information selection by keyword is a person who sees all the target sentences that are the selection results from the selection result list. Need to judge.
In reality, in order for a person to judge a search result, he / she reconfirms the result and judges whether it is correct. The judgment is to confirm whether the search result is what you expected, and whether it is what you expected is easy if the points necessary for the judgment include the procedure that derived the result. You can judge. Therefore, a point that serves as a reference for judgment is arbitrarily set, and this point procedure is accumulated, written, displayed, read aloud, and displayed as an image. Whether or not the search result is correct By checking this procedure text, it is possible to judge whether or not the judgment ground is correct without checking the entire text of the search result. For example, as a point, when the calculation result affects the search result, it is set as a point for documenting the calculation formula and the judgment standard.

FIG. 2 is a diagram showing an example of creating a sentence of a procedure that is the reason / basis of a selection decision and displaying, reading aloud, or displaying an image.
Document important and arbitrarily set judgment points leading to selection as the procedure used for judgment.

Based on the fact that sentences are decomposed into phrases, words in the phrases are extracted, and comparison work is performed, the fact that the contents of the phrases are the same and the relationships between the phrases are the same is the fact that the contents match. , Based on the comparison procedure, and if numbers and units are included, the comparison result determines whether the unit system included in the input text is the same and the numbers are within the comparison range. It is a document creation point where the procedure is set arbitrarily, and is an example of the procedure used as the basis.

Divide the entered text into clauses. The text or sentence group to be selected (regardless of the format that has the text such as a file) is also divided in the same way. Division is based on division with punctuation marks, and clauses that do not contain verbs are combined with clauses that include verbs that follow. Next, the words containing the part of speech are extracted in the phrase. Next, words that are repeated with the same meaning in the phrase are excluded as words that can be compared and excluded. Create a combination that does not differ in meaning even if the order of the words in the phrase is different, and compare each phrase of the input and the selection target. If any criteria that can be judged to be a match are met, it is considered that there is a matching clause, and the content of the clause is judged to be a match. Next, the phrases of the input sentence are rearranged within the range where the meaning does not change, and the combination is written. The same process is performed for the text army to be selected. The phrases are compared from the front of the sentence for all of these phrase combinations, and if the matching phrases are in the same position from the beginning of the sentence, it is judged as a match comparison result. If it contains an arbitrarily set judgment point, add it to the text as a procedure for rationale to reach a conclusion.

If the input text contains numbers, compare the numbers to the selection target. Also, check whether the input text contains units, and if it contains units, determine whether the input text and the unit system of the number to be selected are the same, and if they match, multiply by the unit of the number. Compare the calculated results to determine if the magnitude relationship meets the conditions contained in the text. For example, if the input text describes 10 mm or more and the text to be selected has a description of 5 mm or less, it can be determined that there is a discrepancy. Add this comparison procedure as a procedure to reach the selection result based on the text.

In the case of sentences, for the content matching result that there are clauses with matching contents and the positions where the clauses are arranged match, the comparison procedure is written as the reason / ground for selection. In the case of numbers, the unit to reach the conclusion is similar, and the comparison condition and the comparison process of comparing the size of the numbers are used as the reasons and grounds for selection. For example, by preparing an arbitrary standard document type as shown in Fig. 3 and inserting the comparison procedure and selection result into the standard document, "the content described in clause 2 of input sentence A becomes the content of clause 3 of sentence 2." The reason for selecting the selected procedure by displaying the text, such as "I selected this text because I was able to confirm that they match." It becomes the basis.

Add, display, or read aloud the text that describes the created procedure to the selection result as the reason / ground for selection.

In a system that handles sentences, or a system that handles images and videos, in a system where the explanation of images and videos contains sentences, as a result, two or more pieces of information may be selected, that is, a system that may derive multiple selections as search results. It is available.

Also, in a search system that uses artificial intelligence, there is a concern about reliability if it is possible to grasp whether the search result is correct only by the conclusion, and it can be used in a system that displays it as a basis for proving the correctness of the search result. Is.

Claims (2)

In an information selection support system (so-called search system) that does or does not go through a network, instead of displaying, reading aloud, or displaying an image of the selected result itself or a part of it, a procedure for drawing a conclusion is created and the basis for selection. A system that displays or reads aloud or displays images as a reason A system in which the method, method, means, and logic used in the procedure for drawing a conclusion in claim 1 are documented and displayed, read aloud, or displayed as an image.

Images