JP6770283B1 - Question answer system and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately present answer sentences having different contents in an order which seems to be desired by a user. SOLUTION: In a question-and-answer system 1, a storage unit 31 stores a question-and-answer pattern that is a set of a question-sentence and a corresponding answer-sentence, and a reception-sector 21 accepts input of a question-sentence from a user. The calculation unit 31 calculates the similarity between the question text received by the reception unit 21 and the question text stored in the storage unit 31, and the acquisition unit 31 is based on the similarity calculated by the calculation unit 32. Then, the question-and-answer pattern is acquired from the storage unit 31. When a plurality of question / answer patterns including the same answer sentence exist, the acquisition unit 31 acquires only the question / answer pattern having the highest similarity among them from the storage unit 31, and the presentation unit 22 acquires the question / answer pattern 33. The question-and-answer patterns obtained by are presented to the user in descending order of similarity. [Selection diagram] Fig. 1

Description

In the present invention, a question-and-answer pattern, which is a set of a question sentence and a corresponding answer sentence, is stored in advance, and an appropriate answer sentence is presented from the stored question-and-answer patterns for a new question sentence. It is related to the question and answer system to be asked.

In recent years, natural language processing-type artificial intelligence has become more familiar with the spread of chatbots and smart speakers. For example, in a general keyword search in a text sentence search process, all text sentences including a specified keyword are searched. On the other hand, in the similarity search, which calculates the similarity between a keyword and a text sentence using artificial intelligence and obtains search results based on the similarity, it is possible to rank them in descending order of similarity and provide feedback to the user. It is easier for the user to find the desired content from the search results than the keyword search. However, since Japanese has fluctuations such as "phrase" and "okurigana", the search result may not include the desired content even in the conventional similarity search. There is also a method of using a thesaurus to absorb fluctuations peculiar to Japanese, but maintenance of the thesaurus has a high management load.

For example, in Patent Document 1, an interrogative server that answers questions based on a registered question sentence and an answer sentence to the question sentence corresponds to a question sentence having a high degree of similarity to the question sentence input by the administrator terminal. Display the sentence on the administrator terminal and approve the displayed answer sentence or select another answer sentence to associate the entered question sentence with the approved or selected answer sentence as a new question sentence. A method of memorizing and learning is disclosed. The method described in Patent Document 1 is effective in absorbing fluctuations peculiar to Japanese because it learns so that a plurality of past question sentences are associated with one past answer sentence. It is believed that there is.

Japanese Patent No. 6442807

However, if the past question text is simply increased as in the method described in Patent Document 1, when the user searches, the new question text input by the user corresponds to the same past answer text. The degree of similarity with multiple past question sentences increases. Then, in the method described in Patent Document 1, a question having a certain degree of certainty (= a measure of confidence indicating the same meaning as a new question sentence, and may be similar) is a question having a certain threshold or more. Since the answer patterns are displayed and the question and answer patterns up to a predetermined order are displayed in descending order of certainty, as a result, the displayed answer sentences are all the same. This is inconvenient because if the answer text with the same content is not the desired content, the user will be forced to search again with another question text. Desirably, it is desired to present the answer sentences having different contents in the order in which the user thinks the desired contents.

The present invention has been made in view of the above-mentioned problems, and the purpose of the present invention is to accurately present answer sentences having different contents in the order in which the user thinks that the contents are desired. It is to provide a system and the like.

The first invention for achieving the above-mentioned object is a storage unit that stores a question-and-answer pattern, which is a set of a question sentence and a corresponding answer sentence, and a reception unit that receives input of the question sentence from a user. , The calculation unit for calculating the similarity between the question sentence received by the reception unit and the question sentence stored in the storage unit, and the similarity calculated by the calculation unit. The acquisition unit includes an acquisition unit that acquires the question / answer pattern from the storage unit and a presentation unit that presents the question / answer pattern acquired by the acquisition unit to the user, and the acquisition unit includes the same answer sentence. When there are a plurality of the question-and-answer patterns, only the question-and-answer pattern having the highest degree of similarity is acquired from the storage unit, and the acquisition unit acquires the same question-and-answer pattern including the same answer sentence. When there are a plurality of the same, the question-and-answer pattern from the storage unit is corrected in the order of the corrected degree of similarity by positively correcting the degree of similarity of the question-and-answer pattern having the highest degree of similarity. It is a question-and-answer system characterized by acquiring . According to the first invention, it is possible to prevent the answer sentences presented to the user from having the same contents, and to present a plurality of answer sentences having different contents to the user. Therefore, since the user can confirm the answer sentences of various contents by inputting the question sentence once, it is convenient without being forced to search again by another question sentence. In addition, this makes it possible to accurately present the answer sentences in the order in which the user thinks that the contents are desired.

When making a positive correction, the acquisition unit in the first invention may dynamically change the correction value according to the number of the question / answer patterns including the same answer sentence. This makes it possible to present the question-and-answer patterns with higher accuracy in the order that the user thinks the desired content.

In addition, the acquisition unit in the first invention makes a negative correction to the similarity of all the question-and-answer patterns to be acquired except for the question-and-answer pattern that makes a positive correction to the similarity. You may do so. As a result, the answer sentences can be presented accurately in the order in which the user thinks that the contents are desired.

The second invention was received by a storage unit that stores a question-and-answer pattern, which is a set of a question sentence and a corresponding answer sentence, a reception unit that accepts input of the question sentence from a user, and the reception unit. Based on the calculation unit that calculates the similarity between the question sentence and the question sentence stored in the storage unit, and the similarity calculated by the calculation unit, the question answer pattern is generated from the storage unit. The acquisition unit includes an acquisition unit to be acquired and a presentation unit for presenting the question / answer pattern acquired by the acquisition unit to the user, and the acquisition unit has a plurality of the question / answer patterns including the same answer sentence. In this case, only the question-and-answer pattern having the highest degree of similarity is acquired from the storage unit, and the reception unit is received by the reception unit among the answer sentences presented by the presentation unit. The selection of the answer sentence most suitable for the question sentence is received from the user, and the storage unit sets the question sentence received by the reception unit and the answer sentence received by the reception unit. When the question-and-answer pattern is newly stored, the question-and-answer patterns having the same answer text are given the same label and stored, and the acquisition unit uses the label to store the same answer text. It is a question-and-answer system characterized by determining whether or not there are a plurality of the above-mentioned question-and-answer patterns including. According to the second invention, it is possible to prevent the answer sentences presented to the user from having the same contents, and to present a plurality of answer sentences having different contents to the user. Therefore, since the user can confirm the answer sentences of various contents by inputting the question sentence once, it is convenient without being forced to search again by another question sentence. Moreover, since it is not necessary to compare the answer sentences with each other, the determination process can be executed at high speed.

The third invention is a program for making a computer function as a question-and-answer system, in which the question-and-answer system includes a storage unit that stores a question-and-answer pattern that is a set of a question sentence and a corresponding answer sentence. , A reception unit that accepts the input of the question text from the user, a calculation unit that calculates the degree of similarity between the question text received by the reception unit and the question text stored in the storage unit, and the above. An acquisition unit that acquires the question / answer pattern from the storage unit based on the similarity calculated by the calculation unit, and a presentation unit that presents the question / answer pattern acquired by the acquisition unit to the user. When there are a plurality of question-and-answer patterns including the same answer sentence within a predetermined order from the one with the highest degree of similarity, the acquisition unit has the highest degree of similarity among the question-and-answer patterns. Only the pattern is acquired from the storage unit, and when there are a plurality of the question-and-answer patterns including the same answer sentence, the acquisition unit acquires the same as the question-and-answer pattern having the highest degree of similarity among them. This is a program for making a positive correction every time and causing the question-and-answer pattern to be acquired from the storage unit in descending order of the corrected degree of similarity. By installing the program of the third invention on a general-purpose computer, the question-and-answer system of the first invention can be constructed.

The fourth invention is a program for making a computer function as a question-and-answer system, and the question-and-answer system has a storage unit that stores a question-and-answer pattern, which is a set of a question sentence and a corresponding answer sentence. , A reception unit that accepts the input of the question text from the user, a calculation unit that calculates the degree of similarity between the question text received by the reception unit and the question text stored in the storage unit, and the above. An acquisition unit that acquires the question / answer pattern from the storage unit based on the similarity calculated by the calculation unit, and a presentation unit that presents the question / answer pattern acquired by the acquisition unit to the user. When a plurality of the question-and-answer patterns including the same answer sentence exist, the acquisition unit acquires only the question-and-answer pattern having the highest degree of similarity from the storage unit and receives the reception. The unit receives from the user the selection of the answer sentence most suitable for the question sentence received by the reception unit among the answer sentences presented by the presentation unit, and the storage unit receives the answer sentence by the reception unit. When the question-and-answer pattern, which is a set of the question text and the answer text received by the reception unit, is newly stored, the same label is attached to the question-and-answer patterns having the same answer text. The acquisition unit is a program for assigning and storing, and using the label to make it function so as to determine whether or not there are a plurality of the question-and-answer patterns including the same answer sentence. By installing the program of the fourth invention on a general-purpose computer, the question-and-answer system of the second invention can be constructed.

INDUSTRIAL APPLICABILITY According to the present invention, it is possible to provide a question-and-answer system or the like capable of accurately presenting answer sentences having different contents in an order in which the user thinks that the contents are desired.

The figure which shows the outline of the question answer system 1. The figure which shows an example of the question answer pattern data 5 (5a) Flow chart showing the processing flow of the question-and-answer system 1 Diagram showing an example of the question text reception screen 6 The figure which shows an example of the intermediate data 7 and the acquired data 8. Diagram showing an example of the question / answer pattern presentation screen The figure which shows an example of the question answer pattern data 5 (5b)

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a diagram showing an outline of the question answering system 1. As shown in FIG. 1, the question-and-answer system 1 includes a terminal 2 used by a user (the user includes an administrator of the question-and-answer system 1), and a server 3 connected to the terminal 2 via a network 4. It is a system that accepts a user's question and presents an answer sentence with the content desired by the user.

The terminal 2 includes a reception unit 21 that receives input of a question text from the user. The server 3 has a storage unit 31 that stores a question-and-answer pattern that is a set of a question sentence and a corresponding answer sentence, a question sentence received by the reception unit 21, and a question sentence stored in the storage unit 31. A calculation unit 32 for calculating the similarity of the above, and an acquisition unit 33 for acquiring a question-and-answer pattern from the storage unit 31 based on the similarity calculated by the calculation unit 32. Further, the terminal 2 includes a presentation unit 22 that presents the question / answer pattern acquired by the acquisition unit 33 to the user.

The terminal 2 and the server 3 include a CPU as a control unit (abbreviation of "Central Processing Unit"), a memory as a main storage unit, an HDD as an auxiliary storage unit (abbreviation of "Hard Disk Drive"), a flash memory, and a display unit. A liquid disk display as an input unit, a keyboard or mouse as an input unit, a touch panel display, a LAN (Local Area Network) cable as a wired communication unit, a wireless module as a wireless communication unit, and the like.

The HDD or flash memory as an auxiliary storage unit stores an OS (abbreviation of "Operating System"), an application program, data required for processing, and the like. The CPUs of the terminal 2 and the server 3 read the OS and application programs from the auxiliary storage unit, store them in the main storage unit, control other devices while accessing the main storage unit, and execute the processing described later.

The terminal 2 is, for example, a smartphone, a tablet terminal, a notebook PC (abbreviation of “Personal Computer”), a desktop PC, or the like. The server 3 is, for example, a server computer arranged in a data center or the like, and may be realized by one housing or a plurality of housings. The network 4 is, for example, the Internet, WAN (abbreviation of "Wide Area Network"), or the like.

A program of web browsing software is installed in the auxiliary storage unit of the terminal 2. Further, the program of the web application provided as the question-and-answer system 1 is installed in the auxiliary storage unit of the server 3, and the database necessary for the question-and-answer system 1 is constructed. The terminal 2 can use the web application applied by the server 3 by the function of the web browsing software. A program dedicated to the question-and-answer system 1 may be installed in the auxiliary storage unit of the terminal 2 instead of the program of the web browsing software.

The terminal 2 includes a microphone as a voice input unit and a speaker as a voice output unit. The reception unit 21 accepts the input of a question text from the user by voice, and the presentation unit 22 presents the question and answer pattern to the user by voice. You may. In the following, the reception unit 21 receives a question sentence from the user via the input unit, and the presentation unit 22 presents the question / answer pattern to the user by displaying it on the display unit.

In the above description, the terminal 2 includes the reception unit 21 and the presentation unit 22, and the server 3 includes the storage unit 31, the calculation unit 32, and the acquisition unit 33, but the present invention is not limited thereto. For example, the terminal 2 may include a reception unit 21, a calculation unit 32, an acquisition unit 33, and a presentation unit 22, and the server 3 may include only the storage unit 31. Further, although the question-and-answer system 1 is composed of the terminal 2 and the server 3, the functions of the terminal 2 and the server 3 may be configured as a stand-alone type. Hereinafter, the question-and-answer system 1 will be described as being composed of the terminal 2 and the server 3 shown in FIG.

FIG. 2 is a diagram showing an example of question / answer pattern data 5 (5a). Hereinafter, when comprehensively showing, it is expressed as "question answer pattern data 5", and when concretely showing the examples of FIGS. 2 and 7, it is expressed as "question answer pattern data 5a" and "question answer pattern data 5b". To do.

The question-and-answer pattern data 5 includes a sequence number 51 that is stored in the storage unit 31 of the server 3 and uniquely identifies each, and a question-and-answer pattern that is a set of the question sentence 53 and the corresponding answer sentence 54. Further, in the question / answer pattern data 5, the same label 52 is given to the question / answer patterns having the same answer sentence 54. For example, in the data in which the sequence number 51 of the question / answer pattern data 5a shown in FIG. 2 is "1", the question sentence 53 is "I want to know how to execute the OS Update" and the answer sentence 54 is "From the control panel". Yes, the label 52 is "10001". Further, for example, in the data in which the sequence numbers 51 of the question / answer pattern data 5a shown in FIG. 2 are “2” and “501”, the answer sentences 54 are both “access http: // xyz”. The same "10002" is also given to the label 52.

FIG. 3 is a flowchart showing a processing flow of the question answering system 1. As shown in FIG. 3, the reception unit 21 of the terminal 2 accepts the input of the question text from the user (step S1).

FIG. 4 is a diagram showing an example of the question text reception screen 6. The question text reception screen 6 is a screen displayed on the display unit of the terminal 2, and includes an AI search selection button 61 for accepting search selection by AI (abbreviation of "Artificial Intelligence") and a keyword search selection. Keyword search selection button 62 to accept, clear button 63 to accept deletion of selected contents and input contents, question text input text box 64 to accept input of question text in search by AI, execution instruction of search by AI It has an AI search execution button 65 and the like for accepting. In FIG. 4, it is illustrated by "shading" that the AI search selection button 61 is selected.

When the user inputs a question text into the question text input text box 64 via the input unit of the terminal 2 and presses the AI search execution button 65, the reception unit 21 of the terminal 2 accepts the input of the question text. In the example shown in FIG. 4, the reception unit 21 accepts the sentence "OS Update stops" entered in the question text input text box 64 as the reception question text.

Returning to the description of FIG. Next, when the terminal 2 transmits the reception question text received by the reception unit 21 to the server 3 via the network 4, and the server 3 receives the reception question text, the calculation unit 32 of the server 3 receives the reception question. The degree of similarity between the sentence and the plurality of stored interrogative sentences stored in the storage unit 31 is calculated (step S2).

The calculation unit 32 performs morphological analysis of the reception question sentence and vectorizes it. Specifically, the calculation unit 32 removes stop words of particles, auxiliary verbs, etc. (for example, "ha", "no", "desu", etc.) from the reception question sentence, normalizes words, and converts abbreviations into prototypes. Do and vectorize. Similarly, the calculation unit 32 performs morphological analysis of a plurality of stored interrogative sentences and vectorizes them with the same number of dimensions as the received interrogative sentences.

Then, the calculation unit 32 calculates the degree of similarity between the reception question sentence vector in which the reception question sentence is vectorized and the memory question sentence vector in which the storage question sentence is vectorized. In the present embodiment, the calculation unit 32 calculates the cosine similarity according to the equation (1). The closer the cosine similarity value is to 1, the more similar the vectors are to each other, and the closer the cosine similarity value is to 0, the less similar the vectors are to each other.

Returning to the description of FIG. Next, the acquisition unit 33 of the server 3 acquires the question / answer pattern from the storage unit 31 based on the similarity calculated in step S2 (step S3). In the present embodiment, the acquisition unit 33 corrects the similarity calculated in step S2 in order to accurately acquire the question-and-answer patterns in the order that the user thinks the desired content, and the correction similarity is high. Question-and-answer patterns are acquired from the storage unit 31 in order.

FIG. 5 is a diagram showing an example of intermediate data 7 and acquired data 8. FIG. 5 shows an example in which the similarity is calculated by using the sentence “OS Update stops” shown in FIG. 4 as the reception question sentence and the question sentence 53 of the question answer pattern data 5a shown in FIG. 2 as the memorized question sentence. Is shown.

The intermediate data 7 shown in FIG. 5A includes the calculation result of the similarity by the calculation unit 32 and the correction result of the similarity by the acquisition unit 33, and includes the sequence number 71, the label 72, the similarity 73, and the correction similarity. It has a degree of 74 and. The intermediate data 7 shown in FIG. 5A are arranged in descending order of similarity 73.

When there are a plurality of question / answer patterns including the same answer sentence in a predetermined rank (for example, the top 100) from the one with the highest degree of similarity, the acquisition unit 33 has the highest degree of similarity among the question / answer patterns. Only the pattern is acquired from the storage unit 31. As a result, it is possible to prevent the answer sentences presented to the user from having the same content, and to present a plurality of answer sentences having different contents to the user. Therefore, since the user can confirm the answer sentences of various contents by inputting the question sentence once, it is convenient because the user is not forced to search again with another question sentence.

For example, in the intermediate data 7 in which the sequence number 71 shown in FIG. 5A is “2” and “501”, the label 72 is both “10002”. As described above in the explanation of FIG. 2, since the same label is given to the question and answer patterns having the same answer sentence, the intermediate data 7 in which the sequence numbers 71 are “2” and “501” are the same. This is the answer. Further, since the similarity is 1st and 2nd, it is included in the predetermined order from the one with the highest similarity. Therefore, the acquisition unit 33 does not acquire the question-and-answer pattern in which the sequence number 71 having the highest similarity among these is the question-and-answer pattern of "2", and does not target the question-and-answer pattern in which the sequence number 71 is "501". Similarly, the intermediate data 7 in which the sequence numbers 71 are “502” and “3” are similar among them because the labels 72 are both “1003” and are included in the predetermined order from the one with the highest degree of similarity. The sequence number 71 having the highest degree is the acquisition target for the question / answer pattern of "502", and the question / answer pattern for which the sequence number 71 is "3" is not the acquisition target.

Here, the acquisition unit 33 determines whether or not there are a plurality of question / answer patterns including the same answer sentence by using the label assigned to each question / answer pattern in advance and stored in the storage unit 31. As a result, it is not necessary to compare the answer sentences with each other, so that the determination process can be executed at high speed.

Further, in the present embodiment, when there are a plurality of question / answer patterns including the same answer sentence within a predetermined order from the one with the highest degree of similarity, the acquisition unit 33 asks the question with the highest degree of similarity among them. Make a positive correction to the similarity of the answer patterns. Details will be described later in the explanation of step S3 of FIG. 3, but the answer text that is determined to be the content desired by the user in the past is combined with the reception question text received by the reception unit 21 as a new question answer pattern. It is stored in the storage unit 31. Therefore, it is considered that there is a relatively high possibility that the answer text that is commonly included in the plurality of question and answer patterns is judged to be the desired content even for the user who newly inputs the reception question text. Therefore, by making a positive correction to such a question-and-answer pattern, it is possible to present the question-and-answer pattern with high accuracy in the order that the user thinks the content is desired.

Since the intermediate data 7 having the sequence numbers 71 of “2” and “502” shown in FIG. 5A has the highest degree of similarity among the same labels 72, the acquisition unit 33 makes a positive correction. In the example shown in FIG. 5, the positive correction value is set to "0.03". Therefore, in the intermediate data 7 in which the sequence number 71 is “2”, “0.03” is added to the similarity 73 of “0.89”, and the correction similarity 74 is “0.92”. Further, in the intermediate data 7 in which the sequence number 71 is "502", "0.03" is added to the similarity 73 of "0.68", and the correction similarity 74 is "0.71".

Further, when making a positive correction, the acquisition unit 33 may dynamically change the correction value according to the number of question / answer patterns including the same answer sentence within a predetermined order from the one with the highest similarity. good. For example, the acquisition unit 33 may dynamically change the correction value in proportion to the number of question / answer patterns, or may use a predetermined correction function (the correction function is a linear function) with the number of question / answer patterns as arguments. The correction value may be dynamically changed based on (not limited to). As a result, the number of users who have determined that the content is desired in the past can be reflected in the correction value, so that it is possible to present the question / answer pattern more accurately in the order in which the content seems to be desired by the user. Become.

Further, in the present embodiment, the acquisition unit 33 makes a negative correction to the similarity of all the question / answer patterns to be acquired except for the question / answer pattern that makes a positive correction to the similarity. As described above, the answer sentence determined to be the content desired by the user in the past is stored in the storage unit 31 as a new question-and-answer pattern together with the reception question sentence received by the reception unit 21. Therefore, it is considered that it is relatively unlikely that the answer sentence that is not included in the plurality of question and answer patterns in common is judged to be the desired content for the user who newly inputs the reception question sentence. Therefore, by making a negative correction to such a question-and-answer pattern, it is possible to accurately present the question-and-answer pattern in the order that the user thinks the content is desired.

Since the intermediate data 7 in which the sequence numbers 71 shown in FIG. 5A show “4” and “1” does not have the same label 72 elsewhere, the acquisition unit 33 makes a negative correction. In the example shown in FIG. 5, the negative correction value is set to "0.02". Therefore, the intermediate data 7 in which the sequence number 71 is "4" has the similarity 73 of "0.72" minus "0.02", and the correction similarity 74 is "0.70". Further, in the intermediate data 7 in which the sequence number 71 is "1", "0.02" is subtracted from the similarity 73 of "0.55", and the correction similarity 74 is "0.53".

The acquired data 8 shown in FIG. 5 (b) is the result acquired by the acquisition unit 33 from the storage unit 31 in descending order of the correction similarity 74 of the intermediate data 7 shown in FIG. 5 (a). , Label 82, correction similarity 83, question sentence 84, and answer sentence 85. The question sentence 84 and the answer sentence 85 are question and answer patterns presented to the user. The acquired data 8 does not include a plurality of the same label 82, that is, the same answer sentence 85. Then, the acquired data 8 are arranged in an order that is considered to be the content desired by the user as a response sentence to the reception question sentence that "Update of the OS is stopped" received by the reception unit 21.

Returning to the description of FIG. Next, when the server 3 transmits the question / answer pattern acquired by the acquisition unit 33 to the terminal 2 via the network 4, and the terminal 2 receives the question / answer pattern, the presentation unit 22 of the terminal 2 answers the question. By displaying the pattern on the display unit, it is presented to the user (step S4). The presentation unit 22 displays the question / answer patterns on the display unit in the order acquired by the acquisition unit 33, that is, in descending order of correction similarity.

FIG. 6 is a diagram showing an example of a question / answer pattern presentation screen. The question-and-answer pattern presentation screen 9 is a screen displayed on the display unit of the terminal 2, and displays the input question text display item 91 for displaying the reception question text received by the reception unit 21 and the correction similarity. It has similarity display items 92a to 92d for displaying question and answer patterns, question and answer pattern display items 93a to 93d for displaying question and answer patterns, and optimal selection buttons 94a to 94d for accepting selection of the most suitable answer sentence for the reception question sentence. .. The content of the input question text display item 91 is the same as that of the question text input text box 64 shown in FIG. The contents of the similarity display items 92a to 92d and the question / answer pattern display items 93a to 93d are the same as the correction similarity 83, the question sentence 84, and the answer sentence 85 of the acquired data 8 shown in FIG. 5 (b).

Returning to the description of FIG. Next, the terminal 2 is most suitable for the reception question sentence (= the content of the input question sentence display item 91 shown in FIG. 6) received by the reception unit 21 in step S1 among the answer sentences presented by the presentation unit 22. It is confirmed whether or not the selection of the answer sentence is accepted, that is, whether or not the pressing of any of the optimum selection buttons 94a to 94d is accepted (step S5).

When the selection is accepted (Yes in step S5), the terminal 2 receives the reception question sentence received by the reception unit 21 in step S1 and the reception response sentence received by the reception unit 21 in step S5 via the network 4. Is sent to the server 3. When the server 3 receives the reception question text and the reception response text, the storage unit 31 of the server 3 newly stores the question and answer pattern in which the reception question text and the reception response text are combined (step S6). Here, the storage unit 31 assigns the same label to the question and answer patterns having the same answer sentence and stores them. That is, the storage unit 31 assigns the label 82 of the acquired data 8 corresponding to the reception response sentence and stores it. If the selection is not accepted (No in step S5) or after step S6, the process ends.

FIG. 7 is a diagram showing an example of question / answer pattern data 5 (5b). In the question-and-answer pattern data 5b shown in FIG. 7, data having a sequence number 51 of "1001" is added to the contents of the question-and-answer pattern data 5a shown in FIG. In the example shown in FIG. 7, the reception unit 21 of the terminal 2 accepts the selection of the optimum selection button 94a on the question / answer pattern presentation screen 9 shown in FIG. 6, and the storage unit 31 of the server 3 serves as a new question / answer pattern. , The question sentence 53 "Update of the OS stops", and the answer sentence 54 "access http: // xyz ~" is memorized by adding the label 52 of "10002".

As described above, according to the question / answer system 1, the storage unit 31 stores the question / answer pattern, which is a set of the question text and the corresponding answer text, and the reception unit 21 inputs the question text from the user. The reception / calculation unit 32 calculates the similarity between the question text received by the reception unit 21 and the question text stored in the storage unit 31, and the acquisition unit 33 calculates the similarity degree calculated by the calculation unit 32. The question-and-answer pattern is acquired from the storage unit 31 based on. When there are a plurality of question / answer patterns including the same answer sentence, the acquisition unit 33 acquires only the question / answer pattern having the highest similarity among them from the storage unit 31, and the presentation unit 22 acquires the question / answer pattern 33. The question-and-answer patterns acquired by are presented to the user in descending order of similarity. Therefore, the question-and-answer system 1 can present answer sentences having different contents in the order in which the user thinks that the contents are desired.

When there are a plurality of question / answer patterns containing the same answer sentence in the storage unit 31, the question / answer patterns having a similarity equal to or higher than a certain threshold value are displayed, or the question / answer patterns up to a predetermined rank are displayed in descending order of similarity. As a result, the displayed answer sentences will be the same. This is inconvenient because if the answer text with the same content is not the desired content, the user will be forced to search again with another question text. On the other hand, when there are a plurality of question / answer patterns including the same answer sentence, the acquisition unit 33 of the question / answer system 1 acquires only the question / answer pattern having the highest degree of similarity from the storage unit 31. The user can confirm a plurality of answer sentences having different contents by inputting the question sentence once. Therefore, it is convenient because you will not be forced to search again with another question.

Further, when there are a plurality of question / answer patterns including the same answer sentence, the acquisition unit 33 makes a positive correction to the similarity of the question / answer pattern having the highest similarity among them, and also adds to the similarity. The similarity of all the question and answer patterns to be acquired except for the question and answer pattern to be corrected is negatively corrected, and the question and answer patterns are acquired from the storage unit 31 in descending order of the corrected similarity. .. Therefore, the question-and-answer system 1 can accurately present the answer sentences in the order in which the user thinks that the contents are desired.

Although preferred embodiments of the question-and-answer device according to the present invention have been described above with reference to the accompanying drawings, the present invention is not limited to such examples. It is clear that a person skilled in the art can come up with various modified examples or modified examples within the scope of the technical idea disclosed in the present application, and these also naturally belong to the technical scope of the present invention. Understood.

1 ………… Question and answer system 2 ………… Terminal 3 ………… Server 4 ………… Network 5, 5a, 5b ………… Question and answer pattern data 6 ………… Question text reception screen 7 ………… Intermediate data 8… …… Acquisition data 9 ………… Question and answer pattern presentation screen 21 ………… Reception department 22 ………… Submission department 31 ………… Memory department 32 ………… Calculation department 33 ………… Acquisition department

Claims (6)

A storage unit that stores a question-and-answer pattern, which is a set of a question and a corresponding answer,
The reception section that accepts the input of the question text from the user,
A calculation unit that calculates the degree of similarity between the question text received by the reception unit and the question text stored in the storage unit.
An acquisition unit that acquires the question-and-answer pattern from the storage unit based on the similarity calculated by the calculation unit, and
A presentation unit that presents the question / answer pattern acquired by the acquisition unit to the user,
With
When there are a plurality of the question-and-answer patterns including the same answer sentence, the acquisition unit acquires only the question-and-answer pattern having the highest degree of similarity from the storage unit .
When there are a plurality of the question-and-answer patterns including the same answer sentence, the acquisition unit makes a positive correction to the similarity of the question-and-answer pattern having the highest similarity among them, and corrects the same. A question-and-answer system, characterized in that the question-and- answer patterns are acquired from the storage unit in descending order of similarity . The question and answer according to claim 1 , wherein the acquisition unit dynamically changes the correction value according to the number of the question and answer patterns including the same answer sentence when making a positive correction. system. The acquisition unit claims, characterized in that to perform a negative correction to the similarity of the question-and-answer pattern as the all other acquisition target except for the question answer pattern to perform a positive correction to the similarity 1 or the question and answer system according to claim 2 . A storage unit that stores a question-and-answer pattern, which is a set of a question and a corresponding answer,
The reception section that accepts the input of the question text from the user,
A calculation unit that calculates the degree of similarity between the question text received by the reception unit and the question text stored in the storage unit.
An acquisition unit that acquires the question-and-answer pattern from the storage unit based on the similarity calculated by the calculation unit, and
A presentation unit that presents the question / answer pattern acquired by the acquisition unit to the user,
With
When there are a plurality of the question-and-answer patterns including the same answer sentence, the acquisition unit acquires only the question-and-answer pattern having the highest degree of similarity from the storage unit.
The reception unit receives from the user the selection of the answer sentence most suitable for the question sentence received by the reception unit among the answer sentences presented by the presentation unit.
When the storage unit newly stores the question-and-answer pattern in which the question sentence received by the reception unit and the answer sentence received by the reception unit are combined, the answer sentence is the same. Give the same label to each question and answer pattern and memorize it.
The acquisition unit uses the label to determine whether or not there are a plurality of the question and answer patterns including the same answer sentence.
A question-and-answer system that features this . A program that allows a computer to function as a question-and-answer system.
The question-and-answer system
A storage unit that stores a question-and-answer pattern, which is a set of a question and a corresponding answer,
The reception section that accepts the input of the question text from the user,
A calculation unit that calculates the degree of similarity between the question text received by the reception unit and the question text stored in the storage unit.
An acquisition unit that acquires the question-and-answer pattern from the storage unit based on the similarity calculated by the calculation unit, and
A presentation unit that presents the question / answer pattern acquired by the acquisition unit to the user,
With
When the acquisition unit has a plurality of question-and-answer patterns including the same answer sentence within a predetermined order from the one with the highest similarity, only the question-and-answer pattern having the highest similarity among them exists. Is obtained from the storage unit ,
When there are a plurality of the question-and-answer patterns including the same answer sentence, the acquisition unit makes a positive correction to the similarity of the question-and-answer pattern having the highest similarity among them, and corrects the same. A program for functioning to acquire the question-and-answer pattern from the storage unit in descending order of similarity . A program that allows a computer to function as a question-and-answer system.
The question-and-answer system
A storage unit that stores a question-and-answer pattern, which is a set of a question and a corresponding answer,
The reception section that accepts the input of the question text from the user,
A calculation unit that calculates the degree of similarity between the question text received by the reception unit and the question text stored in the storage unit.
An acquisition unit that acquires the question-and-answer pattern from the storage unit based on the similarity calculated by the calculation unit, and
A presentation unit that presents the question / answer pattern acquired by the acquisition unit to the user,
With
When there are a plurality of the question-and-answer patterns including the same answer sentence, the acquisition unit acquires only the question-and-answer pattern having the highest degree of similarity from the storage unit.
The reception unit receives from the user the selection of the answer sentence most suitable for the question sentence received by the reception unit among the answer sentences presented by the presentation unit.
When the storage unit newly stores the question-and-answer pattern in which the question sentence received by the reception unit and the answer sentence received by the reception unit are combined, the answer sentence is the same. Give the same label to each question and answer pattern and memorize it.
The acquisition unit uses the label to determine whether or not there are a plurality of the question and answer patterns including the same answer sentence.
A program to make it work like this.

Images