Information processing device, method, and program

Abstract

To provide an information processing device, method, and program for selecting an appropriate answer source based on information a user wants to know and providing an answer. [Solution] An information processing device that outputs response information in response to input inquiry information includes a response source selection unit (330) that determines whether the contents of the inquiry information satisfy the conditions for outputting response information, a response unit (340) that generates response information for the inquiry information by referring to information stored in a database when the determination means determines that the conditions are satisfied, and a manned response unit (370) that switches to operator response when the determination means determines that the conditions are not satisfied.

Description

Technical Field

[0001] The present invention relates to an information processing apparatus, method, and program.

Background Art

[0002] Automatic answering technology, so-called bots, has been developed that outputs appropriate answers to questions from users such as customers and in-house staff.

[0003] In particular, those frequently asked by users are also called FAQs (Frequently Asked Questions), and since they can be uniformly answered to assumed questions, they can be efficiently processed by automatic answering by bots.

[0004] For example, Japanese Patent Application Laid-Open No. 2015-36945 (Patent Document 1) discloses a question-and-answer control program including a database storing FAQ information. According to Patent Document 1, it is possible to control the output of answers to questions described in natural language, avoid answers that do not match the intention of the question, and efficiently process the narrowing down of answers.

[0005] However, in Patent Document 1, questions not in the FAQ database cannot be appropriately answered, and the convenience for users is not sufficient. Therefore, further technology for improving the accuracy of answers to user questions has been demanded.

Summary of the Invention

Problems to be Solved by the Invention

[0006] The present invention has been made in view of the problems in the above prior art, and an object thereof is to provide an information processing apparatus, method, and program that select and answer an appropriate answer source based on information that a user wants to know.

Means for Solving the Problems

[0007] That is, according to the present invention, An information processing device that outputs response information based on the input inquiry information, A means for determining whether the content of the inquiry information satisfies the conditions for outputting the response information, If the means for making the determination determines that the conditions are met, the means for generating the response information for the inquiry information by referring to the information stored in the database, If the means for making the determination determines that the conditions are not met, the means for switching to an operator-assisted response. An information processing device is provided, which includes [the following]. [Effects of the Invention]

[0008] As described above, the present invention provides an information processing device, method, and program that select an appropriate source of information based on the information the user wants to know and then provide an answer. [Brief explanation of the drawing]

[0009] [Figure 1] A diagram showing the schematic hardware configuration of the entire system in an embodiment of the present invention. [Figure 2] A diagram showing the hardware configuration included in the information processing device of this embodiment. [Figure 3] A software block diagram included in the information processing device of this embodiment. [Figure 4] A diagram showing examples of the user information database and the response source information database of this embodiment. [Figure 5] A diagram showing an example of the FAQ database of this embodiment. [Figure 6] A flowchart illustrating the process performed by the information processing device of this embodiment. [Figure 7] This figure shows an example of the screen of a terminal used to submit a question in this embodiment. [Figure 8] This figure shows an example of the screen of a terminal used to submit a question in this embodiment. [Figure 9] This figure shows an example of the screen of a terminal used to submit a question in this embodiment. [Figure 10] This figure shows an example of the screen of a terminal used to submit a question in this embodiment. [Figure 11] This figure shows an example of a screen for evaluating responses in this embodiment. [Figure 12] In another embodiment, this figure shows an example of a terminal screen for prompting questions in an interactive format. [Figure 13] This figure shows an example of a screen for setting up a dialogue scenario in another embodiment. [Figure 14] In another embodiment, this figure shows an example of a terminal screen that prompts for questions based on a dialogue scenario. [Modes for carrying out the invention]

[0010] The present invention will be described below with reference to embodiments, but the present invention is not limited to the embodiments described later. In the figures referenced below, the same reference numerals will be used for common elements, and their descriptions will be omitted as appropriate.

[0011] Figure 1 is a diagram showing the schematic hardware configuration of the entire system in an embodiment of the present invention. The system 100 shown in Figure 1 illustrates an environment in which an information processing device 110 and terminals 120 are connected via a network 130 such as the Internet or a LAN. Furthermore, the method of connecting the information processing device 110 and various terminals 120 to the network 130 may be wired communication or wireless communication.

[0012] The information processing device 110 is a device such as a server, which receives user questions sent from the terminal 120 and performs processing to answer those questions.

[0013] The terminal 120 is a device such as, for example, a personal computer 120a, a smartphone 120b, or a tablet 120c, and transmits a question input by a user who is a questioner to the information processing device 110 via the network 130. Further, the terminal 120 includes a screen for displaying an answer or the like transmitted from the information processing device 110, and provides various information to the user.

[0014] Next, the hardware configuration of the information processing device 110 will be described. FIG. 2 is a diagram showing the hardware configuration included in the information processing device 110 of the present embodiment. The information processing device 110 includes a CPU 201, a RAM 202, a ROM 203, a storage device 204, and a communication I / F 205, and each hardware is connected via a bus.

[0015] The CPU 201 is a device that executes a program for controlling the operation of the information processing device 110 and performs predetermined processing. The RAM 202 is a volatile storage device for providing an execution space for the program executed by the CPU 201, and is used for storing and expanding programs and data. The ROM 203 is a non-volatile storage device for storing programs and firmware executed by the CPU 201.

[0016] The storage device 204 is a readable and writable non-volatile storage device that stores an OS, various applications, setting information, various data, etc. for making the information processing device 110 function. Examples of the storage device 204 include an HDD (Hard Disk Drive) and an SSD (Solid State Drive). The communication I / F 205 connects the information processing device 110 and the network 130, and enables communication with other devices via the network 130. Communication via the network 130 may be either wired communication or wireless communication, uses a predetermined communication protocol such as TCP / IP, and can transmit and receive various data.

[0017] The hardware configuration included in the information processing device 110 of this embodiment has been described above. Next, the functional means executed by each piece of hardware in this embodiment will be described with reference to Figure 3. Figure 3 is a software block diagram included in the information processing device 110 of this embodiment.

[0018] The information processing device 110 of this embodiment includes the following functional means: a user authentication unit 310, a question analysis unit 320, an answer source selection unit 330, an answer unit 340, an answer result learning unit 350, an answer result aggregation unit 360, a human-operated support unit 370, and a storage unit 390. In another preferred embodiment, the information processing device 110 may also include a dialogue scenario setting unit 380. The details of each functional means will be described below.

[0019] The user authentication unit 310 is a means of authenticating the login of the user asking the question. For example, by authenticating the user, the user authentication unit 310 can provide answers based on the user's affiliation information, making it possible to select a more appropriate answer to the question.

[0020] The question analysis unit 320 is a means for analyzing the question received from the terminal 120 via the communication interface 205. The question analysis unit 320 performs so-called morphological analysis on the question entered by the user as natural language, and analyzes it as morphemes that make up the question. The question analysis unit 320 notifies the answer source selection unit 330 of the analyzed morphemes.

[0021] The answer source selection unit 330 selects a database to be the source of the answer based on the morphemes analyzed by the question text analysis unit 320. Database selection can be performed by a matching process that compares the morphemes with information such as tags associated with each database. For example, the answer source selection unit 330 determines whether or not to select a database as the source of the answer based on the degree of association between the morphemes and the database's related information tags. The answer source selection unit 330 identifies the database to be the source of the answer by performing the above matching process and determination process for each database. By selecting a database containing relevant words in this way, an appropriate answer can be provided to the user's question.

[0022] Furthermore, if a user has logged in, the response source selection unit 330 may select a database based on each user's information, in addition to the results of the matching process described above. This allows for the selection of a database based on the user's individual information, making it easier to output responses that match the user's intentions.

[0023] The answer unit 340 is a means of responding to a user's question by extracting answer information from a database and notifying the user of the created answer. The answer unit 340 searches the database selected by the answer source selection unit 330 based on the morphemes output by the question text analysis unit 320. More specifically, it extracts answer information to be output as an answer by performing a matching process based on the morphemes and the words contained in the database. Here, the words contained in the database are, for example, words related to the answer information. Note that the format of the FAQ database that serves as the answer source may differ from database to database. Therefore, the answer unit 340 can perform a search according to the format of each answer source database.

[0024] The answer result learning unit 350 is a means of reflecting the results of the answer unit 340's response to the user as learning data in various databases. For example, if the user's question is resolved by the answer, the question and answer are associated and reflected in the database. If the user's question is not resolved, the degree of association between the question and answer is reduced and reflected in the database. By learning the relationship between questions and answers in this way, it becomes easier to obtain appropriate answers to similar questions asked in the future.

[0025] The response result aggregation unit 360 is a means for aggregating the results of the responses given by the response unit 340 to users. For example, the response result aggregation unit 360 can create rankings based on the number of times a question was asked, evaluate whether the user's question was resolved, and aggregate user satisfaction with the answers. By providing the aggregated results to administrators, the response result aggregation unit 360 can perform usage analysis. This allows administrators to configure the system to output more appropriate answers based on the aggregated results.

[0026] The human support unit 370 is a means of answering questions through human support when it is difficult for the bot to provide an answer. For example, if the answer to the user's question is not found in any of the FAQ databases, or if the question cannot be resolved by the bot, the human support unit 370 notifies an operator of the question. The operator who receives the notification can then provide an appropriate answer to the user's question.

[0027] In the above-described embodiment of the information processing device 110, the device is configured to accept questions when a user inputs a question. However, questions may be accepted by other means. For example, in another preferred embodiment, the information processing device 110 can accept questions in a format in which the user and the bot interact via chat (hereinafter referred to as the dialogue format). In this case, the information processing device 110 may include a dialogue scenario setting unit 380. The dialogue scenario setting unit 380 is a means for selecting a database and setting the content of the dialogue for accepting questions. The dialogue scenario setting unit 380 can set up options to present to the user and to present the next option depending on the selected option.

[0028] The storage unit 390 controls the storage device 204 and is a means for storing various databases. The storage unit 390 includes a user information database 391, a source information database 392, and an FAQ database 393. Each database will be explained with reference to Figures 4 and 5. Figure 4 shows examples of the user information database 391 and the source information database 392 of this embodiment, and Figure 5 shows an example of the FAQ database 393 of this embodiment.

[0029] As shown in Figure 4(a), the user information database 391 is configured with associated fields that store an ID to identify the user, the user's department to belong to, and the user's job title. The user authentication unit 310 can obtain various information about the logged-in user by referring to the user information database 391. Furthermore, by obtaining user information, it becomes possible to provide responses corresponding to each user's department to belong to and job title.

[0030] As shown in Figure 4(b), the source information database 392 is configured such that each field stores the name of the FAQ database from which the answer is derived and tag information indicating related terms. For example, the sales department FAQ is associated with keywords such as "on sale" and "specifications," which are expected to be inquired about by users belonging to the sales department. The tag information for the source information database 392 may be set individually, or it may be set in conjunction with the tags of each FAQ database, as described later.

[0031] As shown in Figure 5, the FAQ database 393 is configured with each field associated with a tag related to a question and the answer information. The answer information stored in the FAQ database 393 can, for example, be items related to frequently asked questions (FAQs). For example, if the sales department is frequently asked about product specifications, then, as shown in Figure 5(a), the sales department FAQ database 393a stores the specifications of each product associated with each product name. Alternatively, the sales department FAQ database 393a may also store words related to each product as tag information.

[0032] Similarly, in Figures 5(b), (c), and (d), answer information for frequently asked questions to each department is stored and associated with the department that provides the answer. In particular, in this embodiment, as shown in Figures 5(a) to (d), by providing the storage unit 390 with multiple FAQ databases 393, the answer source selection unit 330 can select the appropriate answer source for each question.

[0033] Note that the configurations of the various databases shown in Figures 4 and 5 are examples only and do not limit the embodiments.

[0034] Up to this point, the functional means included in the information processing device 110 of this embodiment have been described with reference to Figures 3 to 5. The software block described in Figure 3 corresponds to the functional means realized when the CPU 201 executes the program of this embodiment, thereby enabling each piece of hardware to function. Furthermore, the functional means shown in each embodiment may all be realized in software, or some or all of them may be implemented as hardware that provides equivalent functionality.

[0035] Furthermore, the functional means described above do not necessarily have to be included in the configuration shown in Figure 3. For example, in another preferred embodiment, each functional means may be realized through the cooperation of the information processing device 110 and the terminal 120.

[0036] Next, the processes performed by the various functional means described above will be explained. Figure 6 is a flowchart of the processes performed by the information processing device 110 in this embodiment. The information processing device 110 starts processing from step S1000. In step S1001, the process branches depending on whether or not a user operating the terminal 120 has logged in. If there is no login (NO), the process proceeds to step S1003. If there is a login (YES), the process proceeds to step S1002, where the user authentication unit 310 obtains various information about the user from the user information database 391.

[0037] In step S1003, the information processing device 110 receives a question from the user, and the question analysis unit 320 performs morphological analysis on the received question. Then, in step S1004, the answer source selection unit 330 performs matching processing based on the morphemes and related information tags corresponding to each database, and branches the process based on whether or not it is possible for the bot to provide an answer.

[0038] In step S1004, if the bot can provide an answer (YES), proceed to step S1005. On the other hand, if the bot cannot provide an answer (NO), proceed to step S1010 and the answer will be provided by a human. Examples of cases where the bot cannot provide an answer include when the question is too unclear to perform morphological analysis, or when the extracted morphemes are not included in any of the databases.

[0039] Furthermore, if the user is logged in, the response source selection unit 330 may determine whether to adopt each database as a response source not only based on morphemes but also on the user information obtained in step S1002.

[0040] In step S1005, the answer source selection unit 330 determines and selects the FAQ database 393 to be used as the answer source based on the results of the matching process in step S1004. In step S1006, the answer unit 340 sends the answer information extracted based on the matching process between the morphemes and each field of the FAQ database 393 selected as the answer source to the user. The answer unit 340 also sends a notification to the user along with the answer, asking whether the question has been resolved.

[0041] In step S1007, the process branches depending on whether the user's question has been resolved, that is, whether the answer was what the user wanted. If the answer was what the user wanted (YES), the process proceeds to step S1011. On the other hand, if the answer was not what the user wanted (NO), the process proceeds to step S1008.

[0042] In step S1008, the response source selection unit 330 determines whether it is possible to provide an answer using another response source database. If it is not possible to provide an answer using another response source database (NO), the process proceeds to step S1010, and the answer is provided by a human. If it is possible to provide an answer using another response source database (YES), the process proceeds to step S1009, and the answer is continued by the bot. An example of a case where it is possible to provide an answer using another response source database is when there are multiple FAQ databases that were determined to be adopted as response sources in step S1005, and another FAQ database is selected.

[0043] In step S1009, the answer source selection unit 330 selects another FAQ database as the answer source. The information processing device 110 then returns to step S1006. The information processing device 110 repeats the process from step S1006 onward until the user's question is resolved or it is determined that it is difficult for the bot to provide an answer.

[0044] If the response source selection unit 330 determines in steps S1004 or S1009 that it is difficult for the bot to provide an answer, it selects a human responder. If the response source selection unit 330 selects a human responder, the human responder unit 370 notifies an operator of the question in step S1010, and the operator who receives the notification answers the question.

[0045] In step S1011, the answer result learning unit 350 learns the answer results by reflecting the results of the answers given to the user as learning data in various databases. For example, if the bot provides the answer the user needs, the answer result learning unit 350 associates the morphological analysis results of the question with the answer content and stores it in the various databases. This allows the information processing device 110 to learn the answer results and improve the accuracy of answers to subsequent inquiries. The answer result learning unit 350 can also learn about answers other than the answer the user wanted. This allows it to exclude such answers when a similar question is asked in the future, making it easier for the user to get the answer they wanted. Furthermore, the answer result learning unit 350 can also learn about the answers given by operators. This allows it to handle similar questions in the future using bot answers, reducing the burden on operators.

[0046] The process described in Figure 6 allows the system to select the appropriate source of information based on the user's question and provide an answer. Furthermore, because the system can select a source from various databases based on the question, users are more likely to receive an appropriate answer. Additionally, by switching to operator assistance only when the bot cannot handle the request, the number of times operators are required to respond can be reduced, thereby alleviating the burden on operators.

[0047] In the following, this embodiment will be described with more specific examples, referring to Figures 7 to 10. Figures 7 to 10 show examples of the screen of the terminal 120 used to submit a question in this embodiment. In the following description, Figures 4 to 6 will be referred to as appropriate.

[0048] Figure 7(a) shows an example screen displaying the answer to the user's question, "What is the price of the cheapest item in the color display lineup?" (hereinafter referred to as Question A). In this example, when the question analysis unit 320 performs morphological analysis on Question A (step S1003), words such as "color display," "lineup," "cheapest," and "price" are extracted and output as morphemes, as shown in Figure 7(b).

[0049] The response source selection unit 330 refers to the response source information database 392 by morphological matching processing to determine whether or not it is possible for the bot to provide a response (step S1004). If it is possible for the bot to provide a response, it determines whether each database is appropriate as a response source and selects the database to be adopted as the response source (step S1005). One example of a method for selecting a database to be a response source is to determine whether or not the morphemes extracted by morphological analysis are included in the related information tags.

[0050] In the example in Figure 7, referring to the source database 392, the related information tags in the sales department FAQ database contain the word "lineup" extracted from question A. The accounting department FAQ database contains the words "price" and "lowest price," and the product management department FAQ database contains the word "color display." Here, the related information tags in the accounting department FAQ database contain the most extracted morphemes. Therefore, the source selection unit 330 can determine that the accounting department FAQ database is highly likely to be able to answer question A appropriately, and selects the accounting department FAQ database as the source database (step S1005).

[0051] In the example shown in Figure 7, the database from which the response will be obtained is selected by evaluating the number of morphemes contained in the field that stores the related information tags. However, other evaluation methods may also be used. For example, the database could be selected by weighting the words in the related information tags, or by evaluating and selecting based on the learning results from previous responses.

[0052] When the answer source selection unit 330 selects an answer source database, the answer unit 340 searches for an answer to question A by referring to the selected database. In the example in Figure 7, the accounting department FAQ database is selected, so the answer is searched within that database. Here, the product "DisplayA" has the tag "cheapest," so this can be used as the answer to question A. Therefore, the answer unit 340 creates the answer "DisplayA is the cheapest at 100,000 yen" and notifies terminal 120 (step S1006).

[0053] Furthermore, the answer unit 340 may also send a notification such as "Has the problem been solved?" along with the answer, inquiring with the user whether or not the question has been resolved (step S1007). This allows the system to determine whether or not the answer was what the user wanted, and the system can learn the answer content by registering the answer data (step S1011). A specific example of learning answers is when the answer result learning unit 350 adds words and phrases included in the question to the related information tags of the answer source information database 392.

[0054] Next, let's explain the example in Figure 8. Figure 8 shows an example where a user has logged in and the database from which the response will be obtained is selected based on the user information. When a user logs in (step S1001, YES), the user authentication unit 310 retrieves the user's information from the user information database 391 (step S1002). Here, we assume that a user named ID0001, who is a general employee of the sales department, has logged in.

[0055] Figure 8(a) shows an example screen displaying the answer to question B, after a user identified as ID0001 asks "What is the color display lineup?" (hereinafter referred to as question B). In this example, when the question analysis unit 320 performs morphological analysis on question B (step S1003), words such as "color display" and "lineup" are extracted and output as morphemes, as shown in Figure 8(b).

[0056] The response source selection unit 330 refers to the response source information database 392 based on user information and morphemes to determine whether or not it is possible for the bot to provide an answer (step S1004), and if it is possible, it performs a process to select an appropriate database as the response source (step S1005). The selection of the database to be the response source can be done in the same way as in the example of question A in Figure 7, but in the example in Figure 8, the selection can be made taking user information into consideration.

[0057] For example, the sales department's FAQ database includes the related information tag "product lineup," and the product management department's FAQ database includes the related information tag "color display." Therefore, the answer source selection unit 330 can consider these FAQ databases as candidates for answer sources. However, considering user information, since user ID 0001 belongs to the sales department, there is a high probability that question B is a question related to the sales department. Consequently, the answer source selection unit 330 can select the sales department's FAQ database as the answer source database, thereby outputting answer information that matches the intent of the user's question.

[0058] As described above, the answer source selection unit 330 selects the sales department FAQ database as the answer source (step S1005). Then, the answer unit 340 searches for answer information for question B by referring to the selected database and outputs it. In the example in Figure 8, since the question is about the product lineup, the answer unit 340 can create and output an answer such as "We have Display A, Display B, and Display C" (step S1006).

[0059] In this way, by selecting a source of answers based on the question and information about the user, it is possible to provide answers that are closer to the user's intent. Furthermore, the user information may include restrictions on the answers that each user can receive. For example, if the FAQ database 393 that serves as the source of answers contains confidential information, there may be answers that should not be disclosed depending on the user's position. In such cases, the source of answer selection unit 330 can exclude that FAQ database from the list of potential source of answers. In this way, by taking user information into consideration, it is possible to select the FAQ database 393 that serves as the source of answers more appropriately.

[0060] Next, we will explain the examples in Figures 9 and 10. Figures 9 and 10 show examples of screen displays when the user does not receive the desired answer. In Figure 9(a), we assume that a user who wants to know the amount of operating profit asks "How much is the profit?" (hereinafter referred to as Question C).

[0061] In this example, when the question analysis unit 320 performs morphological analysis on question C (step S1003), words such as "profit" are extracted as morphemes and output, as shown in Figure 9(b). The answer source selection unit 330, as in the cases of Figures 7 and 8, refers to the answer source information database 392 to determine whether or not it is possible for the bot to answer (step S1004), and if it is possible, it performs a process to select an appropriate database as the answer source (step S1005).

[0062] Here, let's assume that the accounting department FAQ database containing the phrase "profit margin" as a related information tag is selected. The response unit 340 refers to the accounting department FAQ database, creates an answer regarding "profit," and notifies the user. In the example in Figure 9(a), the result of the response unit 340 answering "The profit margin for color displays is 10%" is displayed on the terminal 120 (hereinafter referred to as answer c1). However, answer c1 is not the answer desired by the user who wants to know the amount of operating profit. Therefore, the user selects "No" in response to the notification "Is this resolved?" displayed on the screen shown in Figure 9(a) (step S1007, NO).

[0063] The response source selection unit 330, upon receiving notification that the user has selected "No," determines whether the answer can be provided using a database other than the currently selected FAQ database (step S1008). In the example in Figure 9, since the related tags in the public relations department FAQ database also include the phrase "net profit," it is determined that the answer can be provided using the public relations department FAQ database (step S1008, YES). The response source selection unit 330 then changes the response source database to the public relations department FAQ database (step S1009).

[0064] The response unit 340, upon receiving notification that the source database for the response has been changed, creates a response by referring to the public relations department's FAQ database and sends it to the terminal 120 (step S1006). Here, let's assume that the response given is, "Net profit for 2018 increased by 15%," as shown in Figure 10(a) (hereinafter referred to as response c2).

[0065] However, since answer c2 is not the answer the user wants to know for their operating profit, the user's question remains unresolved. Therefore, the user selects "No" again in response to the notification "Is it resolved?" (Step S1007, NO).

[0066] The answer source selection unit 330 then determines whether an answer can be provided from a database other than the currently selected FAQ database (step S1008). In the example in Figure 9, since there are no other FAQ databases related to "profit," the answer source selection unit 330 determines that an answer cannot be provided from another database (step S1008, NO). As a result, the answer source selection unit 330 decides to switch to a human operator response, and the human operator response unit 370 notifies the operator of question C and answers c1, c2, etc., and the answer is provided by a human operator (step S1010).

[0067] Figure 10(b) shows an example of a screen displaying the operator's response after switching to human assistance. As shown in Figure 10(b), the operator can provide an appropriate answer by eliciting details of the question and inquiring about the user's intent.

[0068] In this way, if the displayed answer is not what the user wants, the system can make it easier for the user to obtain the desired answer by changing the FAQ database from which the answer is derived or by switching to an operator's response. Note that the switching to an operator's response is not limited to the examples shown above. For example, the system may switch to an operator's response if the number of times the user has selected that the answer is not what they wanted exceeds a threshold. Also, in the case of an unclear question like question C, the answer source selection unit 330 may determine that an answer cannot be provided by the bot (step S1004, NO) and may provide an answer through an operator (step S1010).

[0069] In Figures 7-10, the user, acting as the questioner, is asked to select whether the displayed answer resolved the question in order to learn from the response results. However, a screen like those shown in Figures 11(a) and (b) may also be displayed to allow the user to make a selection. Figure 11 shows an example of a screen for evaluating the answer in this embodiment.

[0070] For example, Figure 11(a) displays the answer to the question along with the message, "Was this helpful?". The user can then evaluate whether the answer was appropriate by rating it as either "Helpful" or "Not Helpful".

[0071] Furthermore, in another preferred embodiment, the user may be asked to select and evaluate the satisfaction level of the answer, as shown in Figure 11(b), for example. In Figure 11(b), the user is asked to evaluate the satisfaction level of the answer on a 5-point scale as an example. By evaluating the answer results numerically in this way, answers that receive a rating higher than a preset threshold can be selected as the target for learning. This makes it possible to efficiently learn the answer results and improve the accuracy of answers to subsequent questions.

[0072] Furthermore, the response result aggregation unit 360 can aggregate the response results obtained as described above and output analysis results as shown in Figure 11(c). Examples of analysis results include, but are not limited to, the number of times the response system was used, the breakdown of user evaluations, the ranking of the number of times the question was asked, and satisfaction levels by question category, as shown in Figure 11(c).

[0073] By reviewing the analysis results, administrators and operators can configure settings to ensure users receive appropriate answers. For example, if a ranking of frequently asked questions reveals items with high question frequency but low user satisfaction with the answers, it becomes possible to modify related information tags or the FAQ database source to provide answers that satisfy users.

[0074] Up to this point, this embodiment has been described with reference to Figures 7-11, illustrating with specific examples. In the example described above, the user's natural language input question was used to select an answer source, but this is only one embodiment, and the information processing device 110 may accept questions in other formats. As an example of a format for accepting questions other than natural language input, questions may be accepted in a dialogue format, such as a chat between the user and the bot. Figure 12 shows an example of the screen of a terminal 120 that asks a question in a dialogue format in another embodiment. In Figure 12, it is assumed that the user wants to know the delivery date of DisplayA.

[0075] In the example dialogue shown in Figure 12, instead of asking questions in natural language, the user identifies the database from which the answer will be obtained by selecting the information they want from the options presented by the answer source selection unit 330. For example, as shown in Figure 12(a), the answer source selection unit 330 notifies the terminal 120 with "What would you like to know about?" and presents the user with options such as "1. Display", "2. Camera", and "3. Other". The user then selects the information they want from the presented options. Here, the user can select an option by typing "Display", or by typing "1", which is the number corresponding to "Display". Alternatively, the user can input by tapping or clicking on the options displayed on the screen with a mouse.

[0076] Upon receiving notification that the user has selected "Display," the Answer Source Selection Unit 330 further presents options related to "Display" and prompts the user to make a selection. In the example in Figure 12(a), it displays "What would you like to know about displays?" and then presents options such as "1. Price" and "2. Specifications." By associating each option with various FAQ databases, the Answer Source Selection Unit 330 can more easily identify the database from which to provide the answer. For example, the "Price" option can be associated with the accounting department's FAQ database, and the "Specifications" option can be associated with the sales department's FAQ database. This allows the Answer Source Selection Unit 330 to appropriately select the FAQ database from which to provide the answer based on the options selected by the user.

[0077] In the example shown in Figure 12(a), none of the presented options relate to "delivery date," which the user wants to know. Therefore, the user enters the "Other" option, as shown in Figure 12(b). If the "Other" option is not associated with any FAQ database, the answer source selection unit 330 determines that a human operator is more appropriate than a bot to answer the user's question and switches to operator support. The operator then receives and answers the user's question, allowing the user to obtain the desired answer. Natural language input may also be accepted during the conversation.

[0078] In this way, even when questions are accepted in a dialogue format, it is possible to appropriately select the source of the answer. Therefore, users are more likely to receive appropriate answers, and the burden on operators is reduced.

[0079] Furthermore, in cases where the user is presented with options and answers are obtained by the user selecting the information they want from among the options, as in the dialogue format shown in Figure 12, it is necessary to repeatedly branch the options to identify the questions. Therefore, the information processing device 110 in another preferred embodiment is configured to include a dialogue scenario setting unit 380. By setting a dialogue scenario (hereinafter referred to as a dialogue scenario) that associates options with answers, it becomes easier to identify the database from which the answers are sourced. Figure 13 shows an example of a screen for setting a dialogue scenario in another embodiment.

[0080] The dialogue scenario setting unit 380 displays a screen, as shown in Figure 13, to an administrator or other user, for example. By operating this screen, the administrator or user can configure the system to present options for greetings and question filtering when a chat between the bot and the user begins.

[0081] In the example in Figure 13, the user is first shown the phrase "How would you like to search?" along with options such as "Search all content," "Refine category search," and "View frequently asked questions." Then, the system is configured to allow the user to further select items related to the option they selected. For example, if the user selects "Refine category search," categories are presented as options, and the user is allowed to select one. Similarly, if the user selects "View frequently asked questions," the user is presented with frequently asked questions, and the user is allowed to select one.

[0082] In this way, by presenting further options based on the user's selection and repeating this process, the user can easily obtain the necessary answer. Furthermore, if the necessary answer cannot be obtained by simply presenting options, the system may switch to natural language input or to an operator-assisted response. This makes it easier for the user to obtain the desired answer. Figure 14 shows an example of the screen of terminal 120 that prompts for a question based on a dialogue scenario in another embodiment.

[0083] When the dialogue scenario is set up as shown in Figure 13, the screen shown in Figure 14(a) is displayed first. If the user selects "Narrow down categories," a screen for selecting categories, as shown in Figure 14(b), is displayed. By selecting options related to the information the user wants to know, the FAQ database from which the answer will be found can be narrowed down, and by repeating this process, the information the user wants to know can be identified.

[0084] Furthermore, if the user selects "View Frequently Asked Questions" on the screen shown in Figure 14(a), the system will display frequently asked questions, as shown in Figure 14(c). This allows the user to directly inquire about the information they want to know.

[0085] In a more preferred embodiment, the above-described dialogue format may be combined with natural language input to accept the input of questions.

[0086] According to the embodiments of the present invention described above, it is possible to provide an information processing device, method, and program that select an appropriate source of information based on the information the user wants to know and then provide an answer.

[0087] Each of the embodiments of the present invention described above can be implemented by a device-executable program written in C, C++, C#, Java®, etc. The program of this embodiment can be stored and distributed on a device-readable recording medium such as a hard disk drive, CD-ROM, MO, DVD, flexible disk, EEPROM, EPROM, etc., and can also be transmitted over a network in a format that can be used by other devices.

[0088] Although the present invention has been described above with reference to embodiments, the present invention is not limited to the embodiments described above. It is included within the scope of the present invention as long as it achieves the effects and advantages of the present invention within the range of embodiments that a person skilled in the art could conceive. [Explanation of Symbols]

[0089] 100...System, 110...Information Processing Device, 120...Terminal, 120a:Personal Computer, 120b:Smartphone, 120c:Tablet, 130...Network, 201...CPU, 202...RAM, 203...ROM, 204...Storage Device, 205...Communication I / F, 310...User Authentication Unit, 320...Question Text Analysis Unit, 330...Answer Source Selection Unit, 340...Answer Unit, 350...Answer Result Learning Unit, 360...Answer Result Aggregation Unit, 370...Human Assistance Unit, 380...Dialogue Scenario Setting Unit, 390...Storage Unit, 391...User Information Database, 392...Answer Source Information Database, 393...FAQ Database [Prior art documents] [Patent Documents]

[0090] [Patent Document 1] Japanese Patent Publication No. 2015-36945

Claims

1. An information processing device that outputs a response in response to inquiry information entered by a user via a terminal, A means for determining whether the content of the inquiry information satisfies the conditions for outputting the response, If the means for making the determination determines that the conditions are met, the means for generating a response to the inquiry information by referring to the response information stored in the database, If the means for making the determination determines that the conditions are not met, the means for switching to an operator-assisted response, In addition to the answer generated by the generating means, a means for notifying the terminal of a message asking whether the answer is the desired answer, so that the user can choose whether or not the answer is the desired answer. Includes, If the user indicates that the answer is not what they wanted, the generating means will generate a new answer by referring to answer information stored in another database. The switching means is characterized in that, in response to a notification by the notification means, the user selects that the new response is not the response desired by the user, and the information processing device then switches to an operator response.

2. The database includes a plurality of pieces of information, each having at least a first field for storing words related to the query information and a second field for storing the answer information. The information processing apparatus according to claim 1.

3. The information processing apparatus according to claim 1 or 2, wherein if the contents included in the inquiry information are not in the database, the determination means determines that the condition is not met.

4. The information processing apparatus according to claim 1, further comprising means for learning that the response is desired when the response to a notification by the notification means is selected as a response desired by the user.

5. The information processing apparatus according to claim 4, wherein the learning means learns that the response is not desired when the response is selected as not desired by the user in response to a notification by the notification means.

6. A method for outputting a response in response to inquiry information entered by a user via a terminal, The steps include determining whether the content of the inquiry information satisfies the conditions for outputting the response, If the determination step determines that the conditions are met, the step of generating a response to the inquiry information by referring to the response information stored in the database, In addition to the answer generated in the generation step, a step of notifying the terminal of a message asking whether the answer is the desired answer, so that the user can choose whether or not the answer is the desired answer. If it is determined in the above determination step that the above conditions are not met, the step of switching to an operator-assisted response, If the user indicates that the answer is not what they wanted, the steps include: generating a new answer by referring to answer information stored in other databases; In the notification step described above, if the user selects that the new response is not the response the user desires, the step also includes switching to an operator-assisted response, Methods that include...

7. A program executed by an information processing device that outputs a response in response to inquiry information entered by a user via a terminal, wherein the information processing device Means for determining whether the content of the inquiry information satisfies the conditions for outputting the response, If the means for making the determination determines that the conditions are met, the means for generating a response to the inquiry information by referring to the response information stored in the database, If the means for making the determination determines that the conditions are not met, the means for switching to an operator-assisted response, In conjunction with the answer generated by the generating means, means for notifying the terminal of a message asking whether the answer is the desired answer, so that the user can choose whether or not the answer is the desired answer. To make it function as, If the user indicates that the answer is not what they wanted, the generating means will generate a new answer by referring to answer information stored in another database. The switching means is characterized in that, in response to a notification by the notification means, if the user selects that the new response is not the response desired by the user, the switching means will also switch to an operator response. program.

Images