JP7172187B2 - INFORMATION DISPLAY METHOD, INFORMATION DISPLAY PROGRAM AND INFORMATION DISPLAY DEVICE - Google Patents

Description

The present invention relates to an information display method, an information display program, and an information display device.

Conventionally, FAQ (Frequently Asked Questions) systems using web pages, chat-type screens, etc. have been used. For example, the FAQ device stores an FAQ list in which questions and answers are associated with each other, and when a question is received from a user on a web page or the like, an answer corresponding to the received question is searched from the FAQ list, and the user to respond to.

In recent years, there is also known a technique of receiving an inquiry sentence such as a keyword from a user and narrowing down FAQ. For example, the FAQ device accepts an inquiry sentence and presents the user with a plurality of question sentences corresponding to keywords extracted from the inquiry sentence. Then, the FAQ device responds with an answer associated with the question selected by the user from among the presented questions.

JP 2014-29668 A Japanese Patent Application Laid-Open No. 2009-48303

However, with the above technology, there are cases where the FAQ cannot be narrowed down efficiently, and correct answers cannot be presented to the user, which may reduce user convenience.

For example, if there are a large number of FAQ questions corresponding to the input query, the effect of narrowing down cannot be obtained, and processing such as prompting the user to re-enter the query frequently occurs, which is rather inefficient. Become. In addition, depending on the screen size of the terminal, the top few of the narrowed down questions may be displayed. burden increases.

It is also conceivable that the FAQ system operator guides the questioner with a talk script or the like and makes the questioner select a category of the FAQ, thereby narrowing down the FAQ. In this case, as many conditional branching scripts as there are categories must be written, and the scripts need to be modified each time the categories are modified.

An object of one aspect of the present invention is to provide an information display method, an information display program, and an information display device capable of efficiently narrowing down information according to a question from a plurality of pieces of information.

In the first plan, the information display method receives a question sentence, refers to a storage unit that stores a plurality of pieces of information each including a question sentence and an answer sentence, and selects the received question among the plurality of pieces of information. Information containing any of the words contained in the sentence is extracted, and based on the appearance status of each of the plurality of words contained in the question sentence contained in the extracted information in the plurality of information, among the plurality of words, One of the words is selected and displayed on the display unit, and when the designation of one of the displayed words is received, the storage unit is referred to, and the designated one of the plurality of information is selected. The computer executes a process of identifying information containing any of the words and displaying an answer sentence included in the identified information on the display unit.

According to one embodiment, it is possible to efficiently narrow down information according to a question sentence from a plurality of information.

FIG. 1 is a diagram illustrating an overall configuration example of a FAQ system according to a first embodiment. FIG. 2 is a diagram for explaining overall processing of the FAQ system according to the first embodiment; FIG. 3 is a diagram for explaining narrowing down of FAQs according to the first embodiment. FIG. 4 is a functional block diagram of the functional configuration of the information display device according to the first embodiment; FIG. 5 is a diagram for explaining an example screen of a chatbot that accepts an inquiry message. FIG. 6 is a diagram for explaining a search result list. FIG. 7 is a diagram for explaining extraction processing. FIG. 8 is a diagram for explaining the processing for calculating the harmonic mean. FIG. 9 is a diagram explaining usefulness of the harmonic mean. FIG. 10 is a diagram explaining the calculation result of the harmonic mean. FIG. 11 is a diagram illustrating generation of a hidden word list in initialization processing. FIG. 12 is a diagram for explaining narrowing down of question sentences in the initialization process. FIG. 13 is a diagram for explaining narrowing down by harmonic mean. FIG. 14 is a diagram for explaining the narrowing down results displayed. FIG. 15 is a diagram for explaining a series of screen display flows. FIG. 16 is a diagram for explaining a series of screen display flows. FIG. 17 is a flowchart illustrating the overall flow of FAQ processing of the FAQ system according to the first embodiment; FIG. 18 is a flowchart illustrating a flow of FAQ narrowing processing according to the first embodiment. FIG. 19 is a flowchart illustrating the flow of word extraction processing according to the first embodiment; FIG. 20 is a diagram for explaining inconsistency of categories due to narrowing down. FIG. 21 is a diagram for explaining narrowing down using a knowledge tree. FIG. 22 is a diagram for explaining the flow of narrowing down using a knowledge tree. FIG. 23 is a diagram for explaining fluctuations of words in a knowledge tree. FIG. 24 is a diagram illustrating an example of absorbing fluctuations in words through dictionary cooperation. FIG. 25 is a diagram illustrating an example of a setting screen. FIG. 26 is a diagram illustrating a hardware configuration example.

Hereinafter, embodiments of an information display method, an information display program, and an information display device disclosed in the present application will be described in detail based on the drawings. In addition, this invention is not limited by this Example. Further, each embodiment can be appropriately combined within a range without contradiction.

[overall structure]
FIG. 1 is a diagram illustrating an overall configuration example of a FAQ system according to a first embodiment. As shown in FIG. 1, in this FAQ system, a questioner terminal 1, an information display device 10, and a FAQ server 50 are connected to communicate with each other via a network regardless of whether it is wired or wireless, and the FAQ from the questioner is received. It is a responsive system.

The questioner terminal 1 is an example of a computer used by the questioner, such as a personal computer or a smart phone. The questioner terminal 1 accesses the FAQ server 50 via the information display device 10 and obtains the answers to the FAQ.

The FAQ server 50 is an example of a computer, such as a server, that stores an FAQ database that is a list of FAQs in which question sentences and answer sentences are associated with each other. The FAQ database stored by the FAQ server 50 associates and stores "id, category ID, category, question text, answer text". The "id" stored here is an identifier that identifies the FAQ, and the "category ID" is an identifier that identifies the category. The “category” indicates the category of the question, the “question” indicates the content of the question asked by the questioner, and the “answer” indicates the content of the answer corresponding to the question. In the example of FIG. 1, the FAQ with the id "100" corresponds to the category "weather" with the category ID "22". to June..." is the associated FAQ.

The information display device 10 is a computer device that obtains from the FAQ server 50 an answer sentence corresponding to a question sentence received from the questioner terminal 1 and responds to the answer sentence, and is, for example, a server. When the information display device 10 receives an inquiry text such as a keyword from the questioner terminal 1, the information display device 10 narrows down the question sentences by retrieving the question sentences including the inquiry sentence in the category or content. Then, the information display device 10 presents a plurality of question sentences obtained by the narrowing down to the questioner terminal 1, acquires an answer sentence corresponding to the question sentence selected by the questioner from the FAQ server 50, and responds.

In such a system, the information display device 10 receives inquiry information related to question sentences. Then, the information display device 10 refers to an FAQ database that stores a plurality of FAQs each including a question sentence and an answer sentence, and selects an FAQ that includes one of the words included in the inquiry information among the plurality of FAQs. Extract. Then, the information display device 10 selects a plurality of words to be selected from the plurality of words based on the harmonic average of each of the plurality of FAQs of the plurality of words included in the question sentences associated with the extracted FAQ. It is displayed on the display or the questioner terminal 1. Subsequently, when the information display device 10 receives the designation of any one of the displayed selection candidates, the information display device 10 refers to the FAQ database and includes the designated word among the plurality of FAQs. Identify the FAQ and display the answers included in the identified FAQ.

Here, the overall flow of the FAQ system will be described using specific numbers as an example. FIG. 2 is a diagram for explaining overall processing of the FAQ system according to the first embodiment; As shown in FIG. 2, the FAQ server 50 stores the FAQ database 51 shown in FIG. 1 (S1). In such a system, the questioner terminal 1 accesses the information display device 10 using a chatbot or the like, and inputs an inquiry sentence (S2 and S3).

Subsequently, the information display device 10 extracts a keyword (hereinafter sometimes referred to as KW) from the inquiry sentence, and obtains a question sentence list 100 containing 10 question sentences corresponding to KW from the FAQ database. (S4 and S5).

Then, the information display device 10 extracts KWs (KW1, KW2, etc.), which are words or the like, from each question in the question list 100, and for each KW, the harmonic mean ( A, B, etc.) are calculated to generate the harmonic mean list 101 (S6). Subsequently, the information display device 10 presents KW candidates 102 including KW1, KW2, and KW3, which are the top three harmonic averages, to the questioner terminal 1 (S7 and S8).

After that, the questioner terminal 1 selects KW1 from the KW candidates 102 presented by the information display device 10 (S9 and S10). Then, the information display device 10 extracts six question sentences including the selected KW1 from each question sentence in the question sentence list 100 and generates a question sentence list 103 (S11).

Next, KWs (KW10, KW11, etc.) excluding the selected KW1 are extracted from each question in the question list 103, and the harmonic average (AA, BB, etc.) to generate the harmonic mean list 104 (S12). Then, the information display device 10 presents KW candidates 105 including KW15, KW18, and KW19, which are the top three harmonic averages, to the questioner terminal 1 (S13 and S14).

After that, the questioner terminal 1 selects KW15 from the KW candidates 105 presented by the information display device 10 (S15 and S16). Then, the information display device 10 extracts three question sentences, namely, question sentences 1, 5, and 9 containing the selected KW18 from the question sentences in the question sentence list 103, and displays them in the question sentence list. 106 is generated (S17).

Here, the information display device 10 presents the question sentence list 106 including the extracted three question sentences to the questioner terminal 1 since the number of question sentences is three below the threshold (S18 and S19). . Then, when the questioner terminal 1 selects the question sentence 5 from the question sentence list 106 (S20 and S21), the information display device 10 acquires the answer sentence 5 corresponding to the question sentence 5 from the FAQ database (S22 and S23), the acquired answer sentence 5 is responded to the questioner terminal 1 (S24 and S25).

Next, the above-described narrowing down process will be specifically described using screen transitions. FIG. 3 is a diagram for explaining narrowing down of FAQs according to the first embodiment. As shown in FIG. 3, the information display device 10 displays a message "What kind of business do you have?" After narrowing down, a KW list 200A including the top three KWs with the highest harmonic mean is displayed (see FIG. 3(a)).

Subsequently, when a "date" is selected from the KW list 200A by the questioner terminal 1 (see (b) of FIG. 3), the information display device 10 displays a question sentence 200b including the selected "date". (Refer to (c) of FIG. 3).

In this manner, the information display device 10 efficiently repeats the selection of keywords with the questioner terminal 1, thereby narrowing down a large number of question sentences. Therefore, the information display device 10 can efficiently narrow down the question sentences by suppressing the event that the question sentences cannot be narrowed down or the event that the question sentences are narrowed down too much.

[Function configuration]
FIG. 4 is a functional block diagram showing the functional configuration of the information display device 10 according to the first embodiment. As shown in FIG. 4 , the information display device 10 has a communication section 11 , a storage section 12 and a control section 20 .

The communication unit 11 is a processing unit that controls communication between the questioner terminal 1 and the FAQ server 50, and is, for example, a communication interface. For example, the communication unit 11 establishes Web communication with the questioner terminal 1 and the FAQ server 50 to transmit and receive data.

The storage unit 12 is an example of a storage device that stores programs and data, such as a memory or a hard disk. This storage unit 12 stores a question sentence list DB 13 and a harmonic mean DB 14 .

The question text list DB 13 is a database that stores question texts included in FAQs selected by keywords input by questioners from the FAQ list stored by the FAQ server 50 . The harmonic average DB 14 is a database that stores the harmonic average of each keyword (word) calculated by the control unit 20 . Details of each DB will be described later.

The control unit 20 is a processing unit that controls the entire information display device 10, such as a processor. The control unit 20 has a reception unit 21 , a search unit 22 and a narrowing processing unit 30 . Note that the receiving unit 21, the searching unit 22, and the narrowing processing unit 30 are an example of an electronic circuit such as a processor or an example of a process executed by the processor. Note that the reception unit 21 is an example of a reception unit.

The reception unit 21 is a processing unit that executes a chatbot between itself and the questioner terminal 1 and receives an inquiry sentence or the like on the chatbot screen. Specifically, upon receiving an access to the FAQ system from the questioner terminal 1, the reception unit 21 transmits a chatbot screen to the questioner terminal 1. FIG. Then, the accepting unit 21 accepts an inquiry sentence including a keyword or the like on the chatbot screen, and outputs the accepted inquiry sentence to the search unit 22 .

FIG. 5 is a diagram for explaining an example screen of a chatbot that accepts an inquiry message. As shown in FIG. 5, the receiving unit 21 displays a chatbot screen 200 on the questioner terminal 1 that has received access, and displays a message such as "What kind of business do you have?" Then, when receiving an inquiry text such as “I want to travel” on the chatbot screen, the reception unit 21 outputs the received inquiry text “I want to travel” to the search unit 22 .

The search unit 22 is a processing unit that uses the inquiry text received by the reception unit 21 to search the FAQ list stored in the FAQ server 50 for the corresponding FAQ question text. Specifically, the search unit 22 performs morphological analysis on the inquiry text to extract keywords, acquires the question text including the keyword from the FAQ list, and stores it in the question text list DB 13 .

For example, the search unit 22 extracts "travel" as a keyword (word) from the query sentence "I want to travel". Then, the search unit 22 searches the FAQ list of the FAQ server for a question including the keyword "travel". FIG. 6 shows the results retrieved in this manner.

FIG. 6 is a diagram for explaining a search result list. As shown in FIG. 6, the search result has "No." and "FAQ question". "No." is an identifier that identifies an FAQ that associates a question and an answer, and "FAQ question" indicates the content of the question. Here, there are 10 questions ranging from "No. 1, I want to go on a sightseeing trip to North America with my family on a cheap plane" to "No. 10, I want to go on a sightseeing trip to North America by myself on an expensive plane." It shows that it was searched. The search unit 22 stores the search results shown in FIG. 6 in the question sentence list DB 13 .

The narrowing-down processing unit 30 has an extraction unit 31, a calculation unit 32, an initialization unit 33, and a display unit 34, and uses the content of the inquiry received from the questioner to specify the FAQ of the questioner. This is a processing unit that executes narrowing down. The narrowing down processing unit 30 successively narrows down the question sentences each time the questioner selects a keyword until the question sentences are narrowed down to a predetermined number. Note that the extraction unit 31 is an example of an extraction unit, and the display unit 34 is an example of a display control unit and an identification unit.

The extraction unit 31 is a processing unit that extracts words and the like that are keywords from each question sentence searched by the search unit 22 . Specifically, the extraction unit 31 uses morphological analysis, predetermined extraction rules, exclusion reasons, an exclusion word list, etc. to extract keywords and the like from each question sentence, and outputs the extraction results to the calculation unit 32. Output.

FIG. 7 is a diagram for explaining extraction processing. As shown in FIG. 7, the extracting unit 31 applies an extraction rule that defines a part of speech to be extracted to the question sentence "image size that can be attached to an email". Specifically, the extraction unit 31 extracts "noun-general" such as "radish" and "acceleration", "noun-proper noun" such as "Hokkaido", Words corresponding to parts of speech such as "adjective-independent" such as "connection" and "medetai" and "verb-independent" such as "tatetsuku" are extracted and an extracted word list is generated.

Here, the extracting unit 31 extracts the form of "prototype"+"nai" when "nai" is added to a word in the imperfect form or the continuous form. For example, the extraction unit 31 extracts words such as "cannot reach" and "cannot connect". The extracting unit 31 also extracts words and compound words from a similar word dictionary stored in advance. For example, the extraction unit 31 extracts "Graduation trip", "How to send mail", "R&D", and the like.

After that, the extracting unit 31 unifies words including compound words included in the extracted word list into vendor expressions. Then, the extracting unit 31 excludes words and the like corresponding to the reason for exclusion from the words and the like whose expressions are unified. For example, the extraction unit 31 excludes a single half-width alphabet, full-width alphabet, or the like from the extraction word list. In addition, the extracting unit 31 excludes from the extracted word list words corresponding to the "noun-sa-hen connection" composed only of symbols such as "*" and "[]" and spaces.

Further, the extraction unit 31 excludes the words registered in the exclusion list from the extracted word list in which the words corresponding to the reasons for exclusion are excluded. For example, the extraction unit 31 excludes words such as "understand, do not understand, come, can, what" registered in the exclusion list. Note that the exclusion list, extraction rules, reasons for exclusion, etc. can be defined in advance and can be updated as appropriate.

In this way, the extracting unit 31 generates an extracted word list including words and the like extracted from the question sentence, and calculates an extracted word list by excluding unnecessary words and the like from the words and the like stored in the extracted word list. 32 and so on. For example, the extracting unit 31 generates an extracted word list including "individual, family, North America, Asia, expensive, cheap, airplane, ship, sightseeing, graduation, go" from the search result of the question sentence shown in FIG. , to the calculation unit 32 . Note that the extraction unit 31 extracts words from each question sentence searched by the search unit 22 and from each question sentence narrowed down by keyword selection by the questioner to generate an extracted word list.

The calculation unit 32 is a processing unit that calculates the harmonic average of each word included in the extracted word list extracted from each question sentence. Specifically, the calculation unit 32 calculates the appearance frequency of each word, etc. included in the extracted word list in all the question sentences that are the extraction source, and uses the calculated appearance frequency to calculate the harmonic average of each word, etc. calculate.

FIG. 8 is a diagram for explaining the processing for calculating the harmonic mean. As shown in FIG. 8, an extraction word list, which is a list of words extracted by the extraction unit 31 from question sentences of each FAQ searched by the search unit 22, is generated. At the time of word extraction, each word is associated with the FAQ. Then, the calculation unit 32 calculates a harmonic average for each word included in the extracted word list, and sorts the calculated harmonic averages in descending order to generate a search result word list. After that, the calculation unit 32 outputs to the display unit 34 or the like a search result word list in which the harmonic averages and the words are associated with each other.

Here, the usefulness of using the harmonic mean instead of the appearance frequency will be explained. FIG. 9 is a diagram explaining usefulness of the harmonic mean. When narrowing down FAQ questions by frequency of appearance of keywords, target questions are less likely to be omitted, but the total number of selection candidates for search target questions cannot be greatly reduced, which is inefficient. In other words, as shown in FIG. 9A, there is a trade-off relationship between the "hardness of omission" and the "easiness of narrowing down" with respect to the appearance frequency of keywords. In a general FAQ system, keywords are narrowed down in order of appearance frequency, so it is possible to prevent candidate question sentences from being omitted from the narrowed down results. , the usefulness of narrowing down cannot be demonstrated.

As described above, there is a trade-off relationship between "difficulty in leaking" and "ease of narrowing down". ” and “difficulty in omission” are well balanced when the keyword frequency is 50%. Therefore, in the narrowing down according to the first embodiment, the "harmonic average" is used to narrow down the FAQs, that is, the question sentences.

Specifically, the calculation unit 32 obtains the search result list shown in FIG. Graduation, go, and get. Then, the calculation unit 32 calculates the frequency of appearance of each word or the like in the extracted word list. For example, the calculation unit 32 calculates the appearance frequency as "80%" because the word "individual" appears in eight of the ten question sentences shown in FIG. Similarly, since the word "family" appears in two of the ten question sentences shown in FIG. 6, the calculation unit 32 calculates the appearance frequency as "20%".

In this way, the calculation unit 32 sets the occurrence frequency of "80% , 20%, 20%, 80%, 40%, 60%, 60%, 40%, 50%, 50%, 100%". Subsequently, the calculation unit 32 substitutes the calculated appearance frequency F into the formula (1), and each word in the extraction word list, such as “individual, family, North America, Asia, expensive, cheap, airplane, ship, sightseeing, Harmonic mean H is calculated for each of "graduation, going". The harmonic mean H takes the maximum value when the appearance frequency F is 50%, and takes the minimum value when it is 100%.

For example, the calculation unit 32 calculates the harmonic mean H ₁ =2/((1/0.8)+(1/(1−0.8)))=0 for “individual” whose appearance frequency is 80%. .32 is calculated. Similarly, the calculation unit 32 calculates the harmonic mean H ₂ =2/((1/0.2)+(1/(1−0.2)))= Calculate 0.32. FIG. 10 shows the results calculated in this manner.

FIG. 10 is a diagram explaining the calculation result of the harmonic mean. As shown in FIG. 10, the calculation unit 32 calculates the harmonic average for each word such as "individual, family, North America, Asia, expensive, cheap, airplane, ship, sightseeing, graduation, go" in the extracted word list as " 0.32, 0.32, 0.32, 0.32, 0.48, 0.48, 0.48, 0.48, 0.50, 0.50, 0.0". Then, the calculation unit 32 outputs the calculation result to the display unit 34 or the like.

Returning to FIG. 4, the initialization unit 33 is a processing unit that executes initialization processing when an inquiry statement is accepted. Specifically, the initialization unit 33 narrows down the FAQ question sentences so that the words included in the inquiry sentence have been selected.

FIG. 11 is a diagram illustrating generation of a hidden word list in initialization processing. As shown in FIG. 11 , the initialization unit 33 executes the same extraction process as the extraction unit 31 on the inquiry text “image size that can be attached to mail” received by the reception unit 21 to obtain the word “ Email, Attachment, Possible, Image, Size" extract. Subsequently, the initialization unit 33 adds the extracted words “mail, attachment, possible, image, size” to the hidden word list.

Subsequently, the initialization unit 33 narrows down the FAQ question sentences. FIG. 12 is a diagram for explaining narrowing down of question sentences in the initialization process. As shown in FIG. 12, the initialization unit 33 extracts words included in both the search result word list calculated by the calculation unit 32 and the hidden word list shown in FIG. 11 to generate a word list. do. Then, the initialization unit 33 extracts question sentences including the words in the word list from the search result list of FAQ question sentences searched by the search unit 22, and generates (updates) a search result list. After that, the initialization unit 33 outputs the search result list to the display unit 34 .

The display unit 34 is a processing unit that displays keywords that are candidates for selection to the questioner. Specifically, the display unit 34 causes the chatbot to display several words with high harmonic averages calculated by the calculation unit 32 as keywords, and allows the questioner to select the corresponding keyword. FIG. 13 is a diagram for explaining narrowing down by harmonic mean. As shown in FIG. 13, the display unit 34 refers to the search result word list, which is a list of harmonic averages calculated by the calculation unit 32, and displays the number of keywords and question sentences containing the keywords in descending order of evaluation average. Extract.

For example, the display unit 34 displays “1, sightseeing, 5,” “2, graduation, 5,” “3, cheap, 4,” “4,” as “refinement candidate display order, keyword, number of question sentences containing keyword”. , airplane, 6”, and so on. Then, the display unit 34 identifies the top three items "1, tourism, 5", "2, graduation, 5", and "3, cheap, 4".

After that, the display unit 34 generates selection candidates using the top three keywords and displays them on the chatbot. FIG. 14 is a diagram for explaining the narrowing down results displayed. As shown in FIG. 14, the display unit 34, in response to the query "I want to travel" from the questioner, selects keywords with high harmonic averages that are predicted to be related, "sightseeing, graduation, cheap" and "none of these." and are displayed as selection candidates.

After that, when "sightseeing" is selected, processing is executed by each processing unit of the narrowing processing unit 30 for the five question statements 1, 2, 3, 4, and 10 including "sightseeing". Then, the newly narrowed down keywords that are new selection candidates are displayed. In addition, when "none of these" is selected, each processing unit of the narrowing down processing unit 30 described above responds to the question sentence that does not include the unselected "sightseeing, graduation, cheap" displayed at that time. is executed to identify new selection candidates.

In this manner, narrowing down and display of selection candidates are repeatedly executed until the number of question sentences containing the selected keyword is equal to or less than a predetermined number. Note that the display unit 34 can also exclude words included in the hidden word list from the search result word list.

[Screen transition]
Next, with reference to FIGS. 15 and 16, a series of screen transitions from the reception of an inquiry to the narrowing down of question sentences will be described. 15 and 16 are diagrams for explaining a series of screen display flows.

As shown in FIG. 15, the information display device 10 displays a chatbot screen on the questioner terminal 1 that has received access, and displays a message "What is your business?" )reference). Subsequently, the information display device 10 receives an inquiry text "I want to travel" on the chatbot screen (see (b) of FIG. 15). Then, the information display device 10 executes the narrowing-down processing by the narrowing-down processing unit 30, and since the number of FAQ question sentences that have been narrowed down is equal to or greater than a threshold value (for example, three cases), the harmonic average is calculated, and based on the harmonic average The selected selection candidates "sightseeing, graduation, cheap, none of these" are displayed (see FIG. 15(c)).

Subsequently, as shown in FIG. 16, the information display device 10 accepts the selection of "sightseeing" from among the selection candidates "sightseeing, graduation, cheap, none of these" (see (d) of FIG. 16). , to identify questions containing the keyword “sightseeing”. Here, since the number of question sentences including the keyword “sightseeing” is two and is less than the threshold, the information display device 10 selects “question sentence A, question sentence B , not in this" is displayed (see FIG. 16(e)). After that, when the "question B" is selected (see (f) in FIG. 16), the information display device 10 refers to the FAQ and displays the "answer B" corresponding to the question B (see FIG. 16(f)). 16 (g)).

In this way, the information display device 10 can narrow down the question sentences, that is, the FAQ through simple exchanges on the chatbot.

[Process flow]
Next, the flow of processing executed by the information display device 10 will be described. Here, the flow of overall processing, the flow of narrowing down processing, and the flow of word extraction processing will be described.

(Overall processing)
FIG. 17 is a flowchart illustrating the overall flow of FAQ processing of the FAQ system according to the first embodiment; As shown in FIG. 17, when the reception unit 21 of the information display device 10 receives access from the questioner terminal 1 (S101: Yes), it displays a chatbot screen and displays a question acceptance message for accepting questions. (S102).

Subsequently, when the search unit 22 detects input of an inquiry sentence on the chatbot screen as a response to the question acceptance message (S103: Yes), it searches the FAQ list based on the inquiry sentence (S104).

Here, if the search result is 0 (S105: Yes), the search unit 22 displays a message indicating that there is no relevant FAQ on the chatbot screen (S106). On the other hand, if the search result is not 0 (S105: No) and is less than the threshold value (S107: Yes), the search unit 22 displays a list of question sentences of each searched FAQ on the chatbot screen (S108 ).

After that, when a desired question text is selected from the list of displayed question texts (S109: Yes), the search unit 22 acquires an answer text corresponding to the selected question text from the FAQ server 50, It is displayed on the bot screen (S110).

On the other hand, in S107, if the search result is not 0 (S105: No) but is equal to or greater than the threshold (S107: No), FAQ narrowing processing is executed (S111).

(Narrowing process)
FIG. 18 is a flowchart illustrating a flow of FAQ narrowing processing according to the first embodiment. Note that this process is a process executed in S111 of FIG.

As shown in FIG. 18, the narrow-down processing unit 30 acquires the question sentence of each FAQ of the search results searched by the search unit 22 (S201). Next, the narrow-down processing unit 30 performs word extraction processing on each of the obtained question sentences (S202).

Thereafter, the narrowing-down processing unit 30 associates the extracted word with the FAQ from which the word was extracted (S203). Subsequently, the narrow-down processing unit 30 calculates the harmonic average for the extracted words, and generates a search result word list sorted in descending order by the harmonic average (S204).

Then, if the query text received for the first time has not yet been processed (S205: No), the narrow-down processing unit 30 acquires the query text input by the questioner (S206). Subsequently, the narrowing down processing unit 30 executes word extraction processing similar to that of S202 (S207). After that, the narrowing down processing unit 30 adds the words extracted from the inquiry sentence to the hidden word list (S208), and extracts FAQs including words existing in both the search result word list and the hidden word list from the search results. (S209).

After executing S209, or when the processing processing for the inquiry text received for the first time in S205 has been completed (S205: Yes), the narrowing-down processing unit 30 adds words contained in the hidden word list to the search result word list. (S210).

Subsequently, the narrow-down processing unit 30 displays the top few words in the word list sorted by the harmonic mean and "Not in this list" as options on the chatbot screen (S211).

After that, when the narrow-down processing unit 30 receives a selection of the displayed option (S212), it executes a process according to the selected option (S213). Specifically, when the keyword "word" is selected from the displayed options, the narrowing down processing unit 30 extracts FAQs including the selected word from the search results (S214), and extracts the selected word. Add to the hidden word list (S215). On the other hand, if "none of these" is selected from the options, the narrow-down processing unit 30 excludes FAQs including words that have been displayed but not selected from the search results (S216).

(word extraction processing)
FIG. 19 is a flowchart illustrating the flow of word extraction processing according to the first embodiment; As shown in FIG. 19, the narrowing-down processing unit 30 performs morphological analysis on input sentences such as inquiry sentences and question sentences to break down the input sentences into words (S301).

Next, the narrow-down processing unit 30 extracts words to be extracted from the decomposed words using a predetermined extraction rule or the like, and generates an extracted word list (S302). Then, the narrowing-down processing unit 30 unifies the extracted words and compound words into vendor expressions (S303).

After that, the narrow-down processing unit 30 excludes one half-width, full-width alphabet from the extracted word list (S304), excludes words composed only of symbols and spaces from the extracted word list (S305), and excludes words registered in the exclusion list. Words that have been extracted are excluded from the extracted word list (S306), and the extracted word list after exclusion is output to each processing unit, the storage unit 12, etc. (S307).

[effect]
As described above, the information display device 10 can display narrow-down keywords in harmonic average order of "difficulty in omission" and "easiness of narrow-down" in order from 50% of frequency, instead of in order of appearance frequency. Therefore, it is possible to suppress inefficient narrowing down, such as when there are candidates that cannot be narrowed down because the candidates are displayed in order of appearance frequency of the narrowing down keyword. As a result, the information display device 10 can suppress the display of candidates that cannot be narrowed down, and can search for the target FAQ with fewer operations than the narrowed-down candidates in order of appearance frequency. can.

By the way, in the narrowing down according to the first embodiment, keywords with low relevance to each other may be displayed as selection candidates. For this reason, the categories of keywords displayed as selection candidates are not unified, and it may be difficult for the questioner to determine which keyword to select.

FIG. 20 is a diagram for explaining inconsistency of categories due to narrowing down. As shown in FIG. 20, the information display device 10 displays keywords in descending order of harmonic average, that is, in descending order of narrowing down efficiency. For this reason, the relevance between keywords, which are candidates for selection, becomes unclear, the information desired by the information display device 10 becomes unclear, and the selection criteria become unclear for the questioner as well.

Therefore, in the second embodiment, a knowledge tree prepared in advance is used to display highly related keywords as selection candidates, thereby improving user convenience. A knowledge tree is a correspondence list in which a category is set as a superordinate concept, keywords belonging to the category are set as a subordinate concept, and the superordinate concept and the subordinate concept are associated with each other.

FIG. 21 is a diagram for explaining narrowing down using a knowledge tree. As shown in FIG. 21, in the knowledge tree, the category "trip purpose" is associated with the keywords "graduation trip" and "sightseeing", and the category "travel destination" is associated with the keywords "Asia" and "America". attached. In such a state, the narrowing-down processing unit 30 of the information display device 10 performs the narrowing-down processing described in the first embodiment for the query sentence "I want to travel", and specifies "sightseeing" as the word with the highest harmonic average. do.

Then, in the second embodiment, the narrowing down processing unit 30 refers to the knowledge tree and acquires entities related to "sightseeing". Specifically, the narrow-down processing unit 30 identifies the category “trip purpose”, which is a superordinate concept of “sightseeing”, and “graduation trip” belonging to the same category. Then, as shown in FIG. 21, the narrowing down processing unit 30 adds the message "Please select the purpose of travel" using the specified category as the title, in addition to "sightseeing" with a high harmonic average, "Graduation trip" to which it belongs is set as an option and displayed to the questioner.

Even if the keywords belong to the same category, keywords that are not included in the question text of the FAQ are not selected. For example, when the searched FAQ and word list are as shown in FIG. 10, even if the keyword "business trip" is included in the category "purpose of travel", the word "business trip" included in the question sentence shown in FIG. , "business trip" is not displayed and is not displayed as an option in FIG.

Here, the flow of narrowing down using the knowledge tree will be described. FIG. 22 is a diagram for explaining the flow of narrowing down using a knowledge tree. As shown in FIG. 22, the narrowing-down processing unit 30 of the information display device 10 narrows down FAQs upon receiving an inquiry sentence "I want to travel".

Then, since the number of relevant FAQ question sentences is equal to or greater than the threshold, the narrowing-down processing unit 30 calculates the harmonic average for the words of each question sentence, and specifies "sightseeing" with the highest harmonic average. After that, the narrow-down processing unit 30 refers to the knowledge tree to identify the keyword "graduation trip" belonging to the same category as the category "travel purpose" of "sightseeing". Then, the narrowing-down processing unit 30 displays selection candidates for selecting the purpose of travel from among the keywords "sightseeing, graduation trip, none of these".

After that, when the questioner selects "sightseeing" from the selection candidates, the narrowing down processing unit 30 searches FAQ question sentences including the word "sightseeing" and narrows down the search.

Then, since the number of relevant FAQ question sentences is equal to or greater than the threshold, the narrowing-down processing unit 30 calculates the harmonic average for the words of each question sentence, and specifies "Asia" with the highest harmonic average. After that, the narrow-down processing unit 30 refers to the knowledge tree to identify the keyword "America" belonging to the same category as the category "travel destination" of "Asia". Then, the narrow-down processing unit 30 displays selection candidates for selecting a travel destination from among the keywords "Asia, America, none of these".

After that, when the questioner selects "Asia" from the selection candidates, the narrowing down processing unit 30 searches FAQ question sentences containing the word "Asia". Then, since the number of FAQ question sentences containing "Asia" is less than the threshold, the narrowing-down processing unit 30 displays each question sentence containing "Asia" as a selection candidate.

In this way, the information display device 10 can narrow down the FAQ questions about "sightseeing" and "Asia" selected by the questioner by repeating questioning until the FAQ questions are narrowed down to the target number. . In other words, the information display device 10 can display the entity that can be narrowed down most efficiently as a selection candidate when asking back a question, and can narrow down the FAQ question sentence with the word selected from the selection candidates and its synonyms. As a result, the information display device 10 can improve user convenience by displaying highly relevant keywords as selection candidates.

By the way, if the knowledge tree used in the second embodiment is used, fluctuations in words may occur, which may actually hinder narrowing down. FIG. 23 is a diagram for explaining fluctuations of words in a knowledge tree. As shown in FIG. 23, "America" is registered as a knowledge tree, and when the questioner selects "America" from the options, an FAQ question sentence containing the word "America" is selected. In this case, question sentences containing "USA" instead of "USA" are not selected. In other words, there are question sentences that are not narrowed down even though they contain words with the same meaning.

Therefore, in the third embodiment, word fluctuations are absorbed by dictionary cooperation. FIG. 24 is a diagram illustrating an example of absorbing fluctuations in words through dictionary cooperation. As shown in FIG. 24, by preparing a dictionary in advance in association with each keyword of the knowledge tree, fluctuations in words are absorbed. For example, the keyword "America" is associated with "America", "United States", "North America", "South America", and "United States".

By doing so, even in the example of FIG. 23, the information display device 10 can narrow down to question sentences including "USA" or "USA" and present them to the questioner. Therefore, narrowing down can be executed efficiently. Information to be set in the dictionary can be arbitrarily changed.

Although the embodiments of the present invention have been described so far, the present invention may be implemented in various different forms other than the embodiments described above.

[Setting screen]
The display examples of the chatbot screen, the message contents, the number of options to be displayed, the threshold for ending the narrowing down, etc. described in the above embodiment are examples, and the settings can be arbitrarily changed on the setting screen or the like.

FIG. 25 is a diagram illustrating an example of a setting screen. As shown in FIG. 25, on the setting screen (1), a threshold for the number of FAQ questions to be narrowed down can be set. For example, to repeat inquiries until the number of FAQ questions reaches 3, set "3". Also, in the area (2) of the setting screen, various settings related to selection candidates to be displayed can be performed. Specifically, the number of keywords to be displayed as selection candidates can be set in (3) of the setting screen. You can set the content of the message to be output as the title of the selection candidate in (4) of the setting screen. You can set the contents to be set in the option for selecting something other than the keyword in (5) of the setting screen.

[Frequency of appearance]
In the above-described embodiment, the example using the harmonic average was explained, but the frequency of occurrence is not limited to this, and can also be used. In addition, the range of appearance frequency (for example, value a to value b) is set in advance so that the “harmonic average” is included in the specific range from value A to value B, and the appearance frequency is set in the range (value a to value b ) can also be extracted.

[Web page]
In the above-described embodiment, an example using a chatbot has been described, but the present invention is not limited to this, and a general web browser, for example, can also be used. Moreover, not only the Internet but also an intranet or the like can be employed. For example, it can be applied not only to FAQs by general users but also to FAQs in the company.

[system]
Information including processing procedures, control procedures, specific names, and various data and parameters shown in the above documents and drawings can be arbitrarily changed unless otherwise specified. Further, the specific examples, distributions, numerical values, etc. described in the examples are merely examples, and can be arbitrarily changed.

Also, each component of each device illustrated is functionally conceptual, and does not necessarily need to be physically configured as illustrated. That is, the specific forms of distribution and integration of each device are not limited to those shown in the drawings. That is, all or part of them can be functionally or physically distributed and integrated in arbitrary units according to various loads and usage conditions. Further, each processing function performed by each device may be implemented in whole or in part by a CPU and a program analyzed and executed by the CPU, or implemented as hardware based on wired logic.

[hardware]
FIG. 26 is a diagram illustrating a hardware configuration example. As shown in FIG. 26, the information display device 10 has a communication device 10a, a HDD (Hard Disk Drive) 10b, a memory 10c, and a processor 10d. 26 are interconnected by a bus or the like.

The communication device 10a is a network interface card or the like, and communicates with other servers and other terminals. The HDD 10b stores programs and DBs for operating the functions shown in FIG.

The processor 10d reads from the HDD 10b or the like a program for executing the same processing as each processing unit shown in FIG. 4 and develops it in the memory 10c, thereby operating processes for executing each function described with reference to FIG. 4 and the like. That is, this process executes the same function as each processing unit of the information display device 10 . Specifically, the processor 10d reads from the HDD 10b or the like a program having functions similar to those of the reception unit 21, the search unit 22, the narrowing processing unit 30, and the like. Then, the processor 10d executes processes for executing processes similar to those of the reception unit 21, the search unit 22, the narrowing processing unit 30, and the like.

In this manner, the information display device 10 operates as an information processing device that executes an information display method by reading and executing a program. Further, the information display device 10 can read the program from the recording medium by the medium reading device and execute the read program, thereby realizing the same function as the embodiment described above. It should be noted that the program referred to in this other embodiment is not limited to being executed by the information display device 10 . For example, the present invention can be applied in the same way when another computer or server executes the program, or when they cooperate to execute the program.

This program can be distributed via a network such as the Internet. Also, this program is recorded on a computer-readable recording medium such as a hard disk, flexible disk (FD), CD-ROM, MO (Magneto-Optical disk), DVD (Digital Versatile Disc), etc., and is read from the recording medium by a computer. It can be executed by being read.

10 information display device 11 communication unit 12 storage unit 13 question sentence list DB
14 Harmonic Mean DB
20 control unit 21 reception unit 22 search unit 30 narrowing processing unit 31 extraction unit 32 calculation unit 33 initialization unit 34 display unit

Claims (7)

a reception process for receiving an inquiry text ;
a plurality of pieces of information retrieved by referring to a storage unit storing a plurality of pieces of information each including a question sentence and an answer sentence, said retrieved plurality of information containing any word included in said inquiry sentence ; an extraction process for extracting a plurality of words from the question sentences included in the retrieved plurality of pieces of information and registering them in a search result word list ;
non-display indicating words included in the query sentence, among the words selected from the plurality of words in the search result word list based on the occurrence status of each of the plurality of words in the retrieved plurality of information; a first display process for displaying on a display unit selection candidates including a predetermined number of words excluding each word in the word list and options indicating words other than the predetermined number of words;
When the designation of any word among the selection candidates is accepted, information containing the designated word among the plurality of information stored in the storage unit is specified, and among the selection candidates, receiving the designation of the option, re-executing the first display process for the information that does not include any of the predetermined number of words, and displaying the selection candidate on the display unit; display processing of
a third display process for displaying an answer text included in the specified information on a display unit;
A method for displaying information, characterized in that the computer executes In the first display process, for each of the plurality of words, the frequency of appearance in question sentences included in the plurality of information retrieved is calculated, and based on the frequency of appearance of each of the plurality of words, calculating the harmonic average of each word, selecting the predetermined number of words having a high calculated harmonic average, and displaying the selection candidates including the selected predetermined number of words and the options on the display unit; The information display method according to claim 1, characterized by: In the first display process, as the predetermined number of words, reference is made to association information that associates a category with a plurality of words belonging to the category, and the first word having the highest harmonic mean is classified into the same category as the first word. 2. A second word belonging to the second word is specified, and the selection candidates including the predetermined number of words including the first word and the second word and the options are displayed on the display unit. 2. The information display method according to 2. The first display processing is characterized in that, among the plurality of second words belonging to the same category, words that are not included in the question sentences included in the plurality of retrieved information are excluded from display targets. The information display method according to claim 3. Determining whether or not the number of retrieved information containing the designated word, which is one of the designated words, is equal to or greater than a threshold;
when the number of the searched information containing the designated word is equal to or greater than the threshold, extracting a plurality of words included in the searched information, and based on the occurrence status of each of the plurality of words in the searched information, selecting one of the plurality of words and displaying it on the display unit;
if the number of the searched information containing the specified word is less than the threshold, each of the searched information is displayed on the display unit, and an answer sentence corresponding to the specified information among the searched information is displayed; 2. The information display method according to claim 1, wherein said computer executes the process of displaying on the display unit. a reception process for receiving an inquiry text ;
a plurality of pieces of information retrieved by referring to a storage unit storing a plurality of pieces of information each including a question sentence and an answer sentence, said retrieved plurality of information containing any word included in said inquiry sentence ; an extraction process for extracting a plurality of words from the question sentences included in the retrieved plurality of pieces of information and registering them in a search result word list ;
non-display indicating words included in the query sentence, among the words selected from the plurality of words in the search result word list based on the occurrence status of each of the plurality of words in the retrieved plurality of information; a first display process for displaying on a display unit selection candidates including a predetermined number of words excluding each word in the word list and options indicating words other than the predetermined number of words;
When the designation of any word among the selection candidates is accepted, information containing the designated word among the plurality of information stored in the storage unit is specified, and among the selection candidates, receiving the designation of the option, re-executing the first display process for the information that does not include any of the predetermined number of words, and displaying the selection candidate on the display unit; display processing of
a third display process for displaying an answer text included in the specified information on a display unit;
An information display program characterized by causing a computer to execute a reception process for receiving an inquiry text ;
a plurality of pieces of information retrieved by referring to a storage unit storing a plurality of pieces of information each including a question sentence and an answer sentence, said retrieved plurality of information containing any word included in said inquiry sentence ; an extraction process for extracting a plurality of words from the question sentences included in the retrieved plurality of pieces of information and registering them in a search result word list ;
non-display indicating words included in the query sentence, among the words selected from the plurality of words in the search result word list based on the occurrence status of each of the plurality of words in the retrieved plurality of information; a first display process for displaying on a display unit selection candidates including a predetermined number of words excluding each word in the word list and options indicating words other than the predetermined number of words;
When the designation of any word among the selection candidates is accepted, information containing the designated word among the plurality of information stored in the storage unit is specified, and among the selection candidates, receiving the designation of the option, re-executing the first display process for the information that does not include any of the predetermined number of words, and displaying the selection candidate on the display unit; display processing of
a third display process for displaying an answer text included in the specified information on a display unit;
An information display device comprising a control unit that executes

Images