JP2019079438A - Answer output program, answer output method, and information processing apparatus - Google Patents

Abstract

To estimate another question to output answer candidates even when there is no answer candidate corresponding to an input question.SOLUTION: An information processing apparatus 101 receives an input of a first question. When there is no answer candidate corresponding to the received first question, the information processing apparatus 101 executes machine learning based on the contents stored in a storage unit 110, and specifies a second question to be input after the received first question. The storage unit 110 stores another question input subsequently when there is no answer candidate corresponding to the question, in association with the question. The other question is a question for which one of answer candidates displayed for the other question is selected. The information processing apparatus 101 outputs answer candidates corresponding to the specified second question.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an answer output program, an answer output method, and an information processing apparatus.

  In recent years, call centers have been increasing to introduce a frequently asked question (FAQ) system in order to level the skills of operators and shorten the response lead time. In addition, the FAQ is disclosed to the outside to reduce the number of inquiries to the call center.

  In the prior art, for example, there is one in which learning of the integrated classifier is performed using the discrimination result and the true class of each of the plurality of classifiers as secondary learning data. In addition, the accuracy of answer candidate extraction of the new learning model information to which additional feature information candidates have been added is compared with the accuracy of answer candidate extraction of the original learning model information, and the accuracy of answer candidate extraction is improved by a predetermined value or more. If there is a problem, there is a technology that keeps the new learning model information in the DB. Further, there is a technique of analyzing history data indicating an operation history, extracting and weighting a keyword included in the history data based on the analysis result, and registering the keyword in the index DB. Also, in the document information that does not include the reference keyword, a keyword that may correspond to the reference keyword is detected, and the average priority of the reference keyword is compared with the corresponding keyword candidate to detect the keyword corresponding to the reference keyword. Have the technology to

JP, 2015-11686, A JP 2007-219995 JP 2005-18530 A JP 2007-199906 A

  However, in the prior art, there are cases where it is not possible to display the answer candidate for the input question.

  In one aspect, the present invention aims to estimate another question and output the answer candidate even if there is no answer candidate corresponding to the inputted question.

  In one embodiment, the input of the first question is accepted, and when there is no answer candidate corresponding to the accepted first question, the other input is accepted subsequently when there is no answer candidate corresponding to the question Machine learning is performed based on the storage content of a storage unit that stores the other question, which is a question and any answer candidate displayed for the other question is selected in association with the question An answer output program is provided, which identifies a second question input following the first question, and outputs answer candidates corresponding to the identified second question.

  According to one aspect of the present invention, even if there is no answer candidate corresponding to the input question, another question can be estimated and the answer candidate can be output.

FIG. 1 is an explanatory view showing an example of the answer output method according to the embodiment. FIG. 2 is an explanatory view showing a system configuration example of the answer output system 200. As shown in FIG. FIG. 3 is a block diagram showing an example of the hardware configuration of the information processing apparatus 101. As shown in FIG. FIG. 4 is an explanatory view showing an example of the storage content of the FAQ master 220. As shown in FIG. FIG. 5 is an explanatory view showing an example of the storage content of the access log DB 230. As shown in FIG. FIG. 6 is an explanatory diagram of an example of the storage content of the teacher data (Yokko) DB 240. As shown in FIG. FIG. 7 is an explanatory diagram of an example of the storage content of the teacher data (bad) DB 250. As shown in FIG. FIG. 8 is an explanatory view showing an example of the storage content of the teacher data (music) DB 260. As shown in FIG. FIG. 9 is an explanatory view showing an example of the FAQ screen 900. FIG. 10 is a block diagram showing an example of a functional configuration of the information processing apparatus 101. As shown in FIG. FIG. 11 is an explanatory diagram (part 1) of a specific example of the access log. FIG. 12 is an explanatory view (part 1) of a specific example of teacher data. FIG. 13 is an explanatory view (part two) of a specific example of the access log. FIG. 14 is an explanatory view (part 2) of a specific example of teacher data. FIG. 15 is an explanatory diagram of an example of the storage content of the analysis DB 1500. FIG. 16 is an explanatory diagram of an example of determination of the FAQ. FIG. 17 is an explanatory view (part 3) of a specific example of the access log. FIG. 18 is an explanatory diagram of an example of time-series change of the first page hit rate. FIG. 19 is an explanatory diagram of an example of a method of estimating morphemes. FIG. 20 is a flowchart (part 1) illustrating an example of a teacher data generation processing procedure of the information processing apparatus 101. FIG. 21 is a flowchart (part 2) of an example of a teacher data generation processing procedure of the information processing apparatus 101. FIG. 22 is a flowchart (part 1) illustrating an example of the response output processing procedure of the information processing apparatus 101. FIG. 23 is a flowchart (part 2) of an example of the response output processing procedure of the information processing apparatus 101.

  Hereinafter, with reference to the drawings, embodiments of an answer output program, an answer output method, and an information processing apparatus according to the present invention will be described in detail.

Embodiment
FIG. 1 is an explanatory view showing an example of the answer output method according to the embodiment. In FIG. 1, the information processing apparatus 101 is a computer that outputs answer candidates corresponding to a question. Questions are for asking how to solve a problem. The question is, for example, a question about a product or service. The question may be expressed by a word or a combination of words, or may be expressed by one or more sentences.

  Also, the answer candidate is information for solving the problem. That is, the answer candidate corresponding to the question is an answer candidate indicating how to solve the question in question. The answer candidate is, for example, an FAQ. The FAQ is an abbreviation of "frequently asked questions" and is a collection of questions and answers in response to questions that are expected in advance.

  Here, in a call center or a support center, a FAQ system is often introduced in order to level the skill of the operator and shorten the answer lead time. Proper operation of the FAQ system leads to an increase in primary response rate and customer satisfaction, so the FAQ system is in an important position in call centers and the like.

  Also, some companies have published their FAQs on the Internet. By publishing the FAQ site outside the company, customers themselves can view the FAQ, so it can be expected to improve customer convenience and reduce the number of inquiries to call centers etc.

  On the other hand, it is desirable to conduct regular review because the contents of the FAQ are easily obsoleted, but the operational load for reviewing the FAQ is large. For example, the number of FAQs registered in the FAQ master may be several thousands to several tens of thousands, and manually reviewing the contents of these FAQs in a short time is very expensive.

  For this reason, the review operation of the FAQ can not follow the name change or new introduction of the product or service, and there is a risk of causing so-called 0 hits in which no one FAQ can be presented for the question. In addition, in the case of accepting a question in a chat format, there is a case where a search is started without confirmation or correction even if there is a typo or conversion error. In such a case, there are often 0 hits.

  If 0 hits occur, the user re-enters the question and re-searches. However, depending on the user, it may be troublesome to re-enter the question, and may give up solving the problem. Therefore, it is important to present some search results for the questions input by the user. On the other hand, outputting a FAQ that has nothing to do with questions does not solve the problem, which may lead to a reduction in the reliability of a call center or the like.

  Therefore, in the present embodiment, an answer output method will be described which estimates another question and outputs an answer candidate even if there is no answer candidate corresponding to the input question. Hereinafter, an exemplary process of the information processing apparatus 101 will be described.

  (1) The information processing apparatus 101 receives an input of the first question. The question to be input may be a word or a combination of words, or may be one or more sentences. The user who inputs the question is, for example, an operator such as a call center or a user who uses the FAQ site. In the example of FIG. 1, it is assumed that the input of the first question "X% #" is received.

  (2) The information processing apparatus 101 determines the presence or absence of an answer candidate corresponding to the received first question. The answer candidate is, for example, an FAQ. The answer candidate corresponding to the question is, for example, a search result (FAQ) for the question, which is searched using an existing search algorithm adopted in the FAQ system or the FAQ site.

  Specifically, for example, the information processing apparatus 101 morphologically analyzes the received first question and breaks it up into morphemes. Next, the information processing apparatus 101 searches the FAQ corresponding to the decomposed morpheme from the FAQ master (for example, the FAQ master 220 shown in FIG. 2 described later) according to a predetermined search condition. Here, when the FAQ is not searched, the information processing apparatus 101 determines that there is no answer candidate corresponding to the first question.

  In the example of FIG. 1, it is assumed that there is no answer candidate corresponding to the received first question "X% #". The information processing apparatus 101 may access another computer to search for an answer candidate corresponding to the received first question.

  (3) If the information processing apparatus 101 determines that there is no answer candidate, the information processing apparatus 101 refers to the storage unit 110 and identifies a second question that is input following the received first question. Specifically, for example, the information processing apparatus 101 performs machine learning based on the storage content of the storage unit 110, and specifies a second question that is input following the first question.

  Here, machine learning is that a computer learns from data, derives patterns and trends, and enables future prediction and decision making. The storage unit 110 stores, in association with the question, another question that has been input and received subsequently when there is no answer candidate corresponding to the question. Here, the other question is a question for which any answer candidate displayed for the other question is selected. That is, the storage unit 110 stores teacher data used for machine learning.

  If it explains in more detail, storage part 110 memorizes the teacher data (music) which matched the other question which continued and accepted the input, when there is no answer candidate corresponding to a question. Further, the other question may be, for example, a question in which any one of the predetermined number N of answer candidates displayed at the top of the other question is selected.

  The predetermined number N can be set arbitrarily. The predetermined number N is set to, for example, the upper limit number of answer candidates displayed on the first page of the screen displaying the answer candidates for the question. In the FAQ system or the FAQ site, the predetermined number N is set to a value of about 30. In the chatbot, the predetermined number N is set to a value of about five.

  In other words, the other question is a question for which one of the answer candidates displayed on the first page of the search result for the other question is inquired (selected) by the user, and the answer for solving the problem is obtained. Is a question that may have been By machine learning such other questions using teacher data associated with the previously input questions, it is a question that is input subsequent to the first question, and the question The feature of the question that is likely to be selected from any of the answer candidates displayed on the first page is derived.

  In the example of FIG. 1, it is assumed that a second question “XXX” input following the first question “X% #” is identified.

  (4) The information processing apparatus 101 outputs the answer candidate corresponding to the specified second question. In the example of FIG. 1, the information processing apparatus 101 searches, for example, the answer candidate corresponding to the identified second question “XXX” from the FAQ master. Here, it is assumed that answer candidates 1 to 5 corresponding to the second question “XXX” are searched. In this case, the information processing apparatus 101 outputs the retrieved answer candidates 1 to 5.

  As described above, according to the information processing apparatus 101, even if there is no answer candidate corresponding to the input first question, the second question is substituted for the first question based on the operation history of the past user. Can be estimated to output answer candidates. The second question is, for example, a question that is input following the first question, and is a question that is highly likely to be selected as any answer candidate displayed for the question. Therefore, even if a zero-hit question is input, it is possible to output an appropriate answer candidate that leads to the solution of the problem.

  The information processing apparatus 101 may store the second question identified in the above (3) in the storage unit 110 in association with the first question. In this case, when the information processing apparatus 101 receives the input of the first question again, the information processing apparatus 101 may specify the second question corresponding to the first question with reference to the storage unit 110. As a result, when a zero-hit question is input again, it is possible to estimate other questions and output answer candidates without performing machine learning.

  In addition, the information processing apparatus 101 may store, in the storage unit 110, the answer candidate corresponding to the second question identified in (3) above in association with the first question. In this case, when the information processing apparatus 101 receives the input of the first question again, the information processing apparatus 101 may refer to the storage unit 110 and output the answer candidate corresponding to the first question. As a result, when a zero-hit question is input again, it is possible to output an answer candidate corresponding to the question without performing machine learning.

  However, with the elapse of time, new teacher data (musical data) is stored in the storage unit 110, and the tendency of other questions subsequently input after a certain question changes. Therefore, it is desirable to specify the second question each time, for example, when it is determined that there is no answer candidate corresponding to the first question. In addition, it is desirable that, for example, the second question stored in the storage unit 110 in association with the first question be periodically reviewed.

(Example of system configuration of answer output system 200)
Next, a system configuration example of the answer output system 200 including the information processing apparatus 101 shown in FIG. 1 will be described. The answer output system 200 may be applied to, for example, an FAQ system such as a call center, or may be applied to an FAQ site published on the Internet.

  FIG. 2 is an explanatory view showing a system configuration example of the answer output system 200. As shown in FIG. In FIG. 2, the answer output system 200 includes an information processing apparatus 101 and a plurality of terminals 201 (three in the example of FIG. 2). In the reply output system 200, the information processing apparatus 101 and the plurality of terminals 201 are connected via a wired or wireless network 210. The network 210 is, for example, a LAN (Local Area Network), a WAN (Wide Area Network), the Internet, or the like.

  Here, the information processing apparatus 101 outputs the answer candidate corresponding to the question. In the following description, “FAQ” may be described as an example of an answer candidate. Further, the information processing apparatus 101 includes an FAQ master 220, an access log DB (Database) 230, a teacher data (Yokoko) DB 240, a teacher data (Rikoko) DB 250, and a teacher data (Magogo) DB 260. The information processing apparatus 101 is, for example, a server.

  The stored contents of the various DBs 220, 230, 240, 250, 260, etc. will be described later with reference to FIGS.

  The terminal 201 is a computer used by the user of the answer output system 200. The user of the answer output system 200 is, for example, a call center operator or a user of an FAQ site. The terminal 201 is, for example, a PC (Personal Computer), a smartphone, a tablet PC, or the like.

(Hardware configuration example of the information processing apparatus 101)
FIG. 3 is a block diagram showing an example of the hardware configuration of the information processing apparatus 101. As shown in FIG. In FIG. 3, the information processing apparatus 101 includes a central processing unit (CPU) 301, a memory 302, an interface (I / F) 303, a disk drive 304, and a disk 305. Also, each component is connected by a bus 300.

  Here, the CPU 301 is in charge of controlling the entire information processing apparatus 101. The memory 302 includes, for example, a read only memory (ROM), a random access memory (RAM), and a flash ROM. Specifically, for example, a flash ROM stores an OS (Operating System) program, the ROM stores an application program, and the RAM is used as a work area of the CPU 301. The program stored in the memory 302 is loaded into the CPU 301 to cause the CPU 301 to execute coded processing.

  The I / F 303 is connected to the network 210 through a communication line, and is connected to an external computer (for example, the terminal 201 illustrated in FIG. 2) through the network 210. The I / F 303 controls the interface between the network 210 and the inside of the device, and controls input / output of data from an external computer. For example, a modem or a LAN adapter can be adopted as the I / F 303.

  The disk drive 304 controls the reading / writing of the data with respect to the disk 305 according to control of CPU301. The disk 305 stores data written under control of the disk drive 304. Examples of the disk 305 include a magnetic disk and an optical disk.

  The information processing apparatus 101 may include, for example, a solid state drive (SSD), an input device, a display, and the like in addition to the above-described components. The terminal 201 illustrated in FIG. 2 can also be realized by the same hardware configuration as the information processing apparatus 101. However, the terminal 201 includes an input device, a display, and the like in addition to the components described above.

(Memory contents of various DB 220, 230, 240, 250, 260, etc.)
Next, storage contents of the various DBs 220, 230, 240, 250, 260, etc., which the information processing apparatus 101 has, will be described with reference to FIGS.

  FIG. 4 is an explanatory view showing an example of the storage content of the FAQ master 220. As shown in FIG. In FIG. 4, the FAQ master 220 has fields of FAQ numbers, questions and answers, and stores the FAQs (for example, FAQs 400-1 and 400-2) as records by setting information in each field.

  Here, the FAQ number is an identifier that uniquely identifies the FAQ. The "#" in the FAQ 400- # corresponds to the FAQ number. The questions are questions that are expected in advance. The answer is an answer to a question that is expected in advance. For example, in FAQ 400-1, the question "Please tell me how to return the PC to its initial state." And the answer "How to recover, etc., when a problem occurs and it does not work properly is as follows. "..." is summarized.

  FIG. 5 is an explanatory view showing an example of the storage content of the access log DB 230. As shown in FIG. In FIG. 5, the access log DB 230 has fields of session number, date and time, type, search word, inquiry FAQ, ranking and search list. By setting information in each field, access logs, for example, access logs 500-1 to 500-3 are stored as records.

  Here, the session number is an identifier that uniquely identifies a session between the information processing apparatus 101 and the terminal 201. A session is a logical connection relationship set up to exchange information in communication between two points (between devices). The date is the date when the operation was performed on the terminal 201.

  The type is the type of operation performed at the terminal 201. Types include, for example, search, query, and the like. The type "search" represents an operation of searching for the FAQ corresponding to the search word. The type “query” represents an operation of selecting a FAQ for a search word, ie, an operation of querying the contents of the FAQ.

  The search word corresponds to the question input at the terminal 201. The search word may be a word or a combination of words, or may be one or more sentences. The inquiry FAQ is the FAQ that has received the selection operation for the search word, that is, the FAQ number of the inquiry FAQ.

  The rank is a rank indicating the number of the FAQ displayed from the top when the FAQ is displayed for the search word. The search list is a list of FAQ numbers of FAQs searched from the FAQ master 220 (see, for example, FIG. 4) based on a search word.

  FIG. 6 is an explanatory diagram of an example of the storage content of the teacher data (Yokko) DB 240. As shown in FIG. In FIG. 6, the teacher data (Yosuke) DB 240 has fields of date and time, inquiry FAQ, and search word after morphological analysis. By setting information in each field, teacher data (for example, teacher data) 600-1 and 600-2 are stored as records.

  Here, the date is the date when the inquiry FAQ is referred to. The inquiry FAQ is the FAQ that has received the selection operation when displayed on the first page for the search word, that is, the FAQ number of the inquiry FAQ. The FAQ displayed on the first page is the top N FAQs among the FAQs for the search word. The morphologically analyzed search word is one or more morphemes detected by morphologically analyzing the search word when the query FAQ is queried.

  FIG. 7 is an explanatory diagram of an example of the storage content of the teacher data (bad) DB 250. As shown in FIG. In FIG. 7, the teacher data (poor) DB 250 has fields of date and time, non-inquiry FAQ, and search word after morphological analysis. By setting information in each field, teacher data (for example, teacher data) 700-1 to 700-3 are stored as records.

  Here, the date and time is the date and time when the non-inquiring FAQ was not referred to, for example, the date and time when the session between the information processing apparatus 101 and the terminal 201 was disconnected without the non-inquiring FAQ being referred to. The non-inquiry FAQ is the FAQ that did not accept the selection operation when displayed on the first page for the search word, that is, the FAQ number of the FAQ that was not inquired. The morpheme-analyzed search word is one or more morphemes detected by morphologically analyzing the search word when the non-inquiring FAQ is not queried.

  FIG. 8 is an explanatory view showing an example of the storage content of the teacher data (music) DB 260. As shown in FIG. In FIG. 8, the teacher data (data base) DB 260 has fields of date and time, alternative word and search word. By setting information in each field, teacher data (music), for example, teacher data (music) 800-1 is stored as a record.

  Here, the date is the date when no FAQ was searched from the FAQ master 220 for the search word, for example, the date when the operation for searching the FAQ corresponding to the search word is performed. The alternative word is another search word that is input following a search word for which no FAQ has been displayed, and any of the FAQs displayed on the first page for the other search word Is the selected search word. The search word is a question for which no FAQ has been searched from the FAQ master 220. The search word may be expressed, for example, by one or more morphemes detected by morphological analysis of the search word.

(Example screen of FAQ screen 900)
Next, a screen example of the FAQ screen 900 displayed on the terminal 201 will be described. In the following description, a case where a click operation using an input device (not shown) is performed as an operation for the user to select a box, button or the like of the operation screen displayed on the terminal 201 will be described as an example.

  FIG. 9 is an explanatory view showing an example of the FAQ screen 900. In FIG. 9, a FAQ screen 900 is an example of an operation screen for searching for a FAQ or inquiring about a FAQ. In the FAQ screen 900, when a box 901 is clicked, a search word can be input. In the example of FIG. 9, the search word "FJ10 initialization" is input.

  Also, when the button 902 is clicked on the FAQ screen 900, the FAQ corresponding to the search word input in the box 901 can be searched. Specifically, when the button 902 is clicked, the search word input in the box 901 is transmitted from the terminal 201 to the information processing apparatus 101.

  As a result, the search result for the search word is transmitted from the information processing apparatus 101 to the terminal 201, and the searched FAQs are listed and displayed in the display area 910. At this time, the top N FAQs are displayed on the first page. Other FAQs will be displayed on page 2 and later. N can be arbitrarily set, and is set to a value of about 5 to 30. Here, N is “N = 5”.

  In the example of FIG. 9, as a result of searching for the FAQ corresponding to the search word “FJ10 initialization”, the top five FAQs 91 1 to 915 are displayed in the display area 910 as the search result of the first page. When the button 903 is clicked on the FAQ screen 900, the search result of the second page can be displayed.

  In addition, when any FAQ displayed in the display area 910 is clicked on the FAQ screen 900, the contents of the FAQ can be displayed. For example, clicking on the FAQ 911 can inquire the contents of the FAQ 911 (questions and answers). When one of the FAQs displayed in the display area 910 is clicked, a selection result indicating that the FAQ is selected is transmitted from the terminal 201 to the information processing apparatus 101.

(Example of functional configuration of the information processing apparatus 101)
FIG. 10 is a block diagram showing an example of a functional configuration of the information processing apparatus 101. As shown in FIG. In FIG. 10, the information processing apparatus 101 includes a reception unit 1001, a search unit 1002, a determination unit 1003, an output control unit 1004, a generation unit 1005, a determination unit 1005, a specification unit 1007, and a storage unit 1010. ,including. The reception unit 1001 to the identification unit 1007 are functions as a control unit, and more specifically, for example, by causing the CPU 301 to execute a program stored in a storage device such as the memory 302 or the disk 305 illustrated in FIG. Alternatively, the function is realized by the I / F 303. The processing result of each functional unit is stored, for example, in a storage device such as the memory 302 or the disk 305.

  The storage unit 1010 is realized by a storage device such as the memory 302 and the disk 305. For example, the storage unit 1010 stores various DBs 220, 230, 240, 250, 260, and 1900 (see FIG. 19 described later) and the like. Note that the storage unit 1010 may be included in another computer different from the information processing apparatus 101. In this case, the information processing apparatus 101 can access another computer and refer to the storage content of the storage unit 1010.

  The receiving unit 1001 receives an input of a question. The question is a search word for asking for a solution to any problem, and may be a word or a combination of words, or one or more sentences. The input of the question is performed, for example, on the FAQ screen 900 shown in FIG. Specifically, for example, the receiving unit 1001 receives an input of a search word by receiving from the terminal 201 the search word input to the FAQ screen 900. In the following description, a question may be described as a "search word".

  The search unit 1002 searches the FAQ corresponding to the search word. Specifically, for example, the search unit 1002 morphologically analyzes the search word and breaks it up into morphemes. Next, the search unit 1002 searches the FAQ master 220 shown in FIG. 4 for the FAQ corresponding to the decomposed morpheme according to the predetermined search condition.

  More specifically, for example, the search unit 1002 may search the FAQ from the FAQ master 220 by setting up an AND condition or an OR condition for a detected morpheme by morphological analysis of the search word. At this time, the search unit 1002 searches for a FAQ with a large number of accesses or a FAQ with a high degree of similarity, taking into consideration, for example, the number of accesses (number of queries) in the past for each FAQ and the similarity with search words. It is also good.

  Thereby, the FAQ corresponding to the search word can be specified. The search result is stored, for example, in the access log DB 230 shown in FIG. 5 as an access log of type “search”. Specifically, the session number of the session between the terminal 201 that received the search word and the information processing apparatus 101 is set as the session number of the access log. The date and time when the search word is accepted is set as the date and time of the access log. The received search word is set in the search word of the access log. Access log query FAQ and ranking is "blank". In the access log search list, the FAQ numbers of the searched FAQs are set. However, when no FAQ corresponding to the search word is searched, “− (null)” is set in the search list of the access log.

  The determination unit 1003 performs machine learning based on the storage content of the storage unit 1010 that stores teacher data (Yokoko) and teacher data (poor), and selects one or more of a plurality of FAQs searched out. Determine multiple FAQs. Here, the teacher data (Yosukeko) is information in which the FAQ in which the selection operation for the question has been received is associated with the question. The teacher data (bad question) is information in which the FAQ for which the selection operation for the question has not been received is associated with the question.

  Specifically, for example, with reference to the teacher data (Yokko) DB 240 shown in FIG. 6, the determination unit 1003 performs machine learning using the naive Bayesian classifier, and performs the first search for each of the plurality of FAQs retrieved. Calculate the probability of The first probability is the probability that the FAQ is selected when the FAQ is displayed for the search word for which the input has been received.

  Thereby, each of a plurality of searched FAQs represents an index value for determining the likelihood of being referred to when displayed on the first page, in other words, the strength of the relevancy with the search word. An index value can be obtained.

  Further, the determination unit 1003 performs machine learning by the naive Bayesian classifier with reference to the teacher data (bad data) DB 250 shown in FIG. 7, and calculates a second probability for each of the plurality of FAQs searched. Do. The second probability is the probability that the FAQ is not selected when the FAQ is displayed for the search word for which the input has been received.

  In this way, each of the plurality of FAQs searched is an index value for determining the possibility of not being inquired when it is displayed on the first page, in other words, an index indicating the weakness of the relevance to the search word You can get a value.

  Then, the determination unit 1003 selects one or more FAQs for the search word from among the searched FAQs based on the first probability and the second probability calculated for each of the searched FAQs. decide. This makes it possible to narrow down the FAQs displayed for the search word in consideration of both the first probability and the second probability.

  Describing in more detail, for example, the determination unit 1003 sorts one or more FAQs obtained by excluding FAQs whose second probability is a threshold α or more among the plurality of searched FAQs in descending order of the first probability. Determine the FAQ as one or more FAQs for the search word. The threshold value α can be set arbitrarily. The threshold α is set to a value of about 80, for example, when the unit is [%].

  Also, for example, the determination unit 1003 selects the top N most frequently used FAQs having a high first probability, and selects the FAQs having the second probability of the threshold α or more among the selected FAQs into one or more FAQs for the search word It may be decided. That is, the determination unit 1003 narrows down the top N FAQs with high first probability, and then excludes FAQs with the second probability equal to or higher than the threshold α.

  The specific processing contents for calculating the first and second probabilities will be described later with reference to FIG. In addition, an example of determination of the FAQ for the search word will be described later with reference to FIG.

  The output control unit 1004 outputs one or more FAQs for the search word. Specifically, for example, the output control unit 1004 outputs the determined one or more FAQs to the search word. Each FAQ is provided with information that can identify the order (display order) of each FAQ. More specifically, for example, the output control unit 1004 transmits one or a plurality of determined FAQs to the terminal 201 to transmit one or more of the search words received in the display area 910 of the FAQ screen 900. Control to display multiple FAQs.

  Further, the receiving unit 1001 receives an operation of selecting a FAQ for a search word. The selection operation of the FAQ is an operation for querying the contents of the FAQ, and is performed, for example, on the FAQ screen 900 shown in FIG. Specifically, for example, the reception unit 1001 receives the selection operation of the FAQ by receiving from the terminal 201 the selection result indicating the FAQ selected (clicked) on the FAQ screen 900.

  When the selection operation of the FAQ is received, an access log of type “inquiry” is stored in the access log DB 230. Specifically, the session number of the session between the terminal 201 that received the search word and the information processing apparatus 101 is set as the session number of the access log. As the date and time of the access log, the date and time when the selection operation of the FAQ is accepted is set. The received search word is set in the search word of the access log. In the access log inquiry FAQ, the FAQ number of the FAQ that has received the selection operation is set. In the order of the access log, the order of the FAQ having received the selection operation is set. Access log search list is "blank". Thereby, it is possible to accumulate an access log indicating the FAQ for which the selection operation has been received when displayed for the search word.

  The generation unit 1005 generates teacher data (Yoshiko). In addition, the generation unit 1005 generates teacher data (bad data). Also, the generation unit 1005 generates teacher data (music). Here, the teacher data (Magogo) is information in which another search word, which is subsequently accepted when there is no FAQ corresponding to the search word, is associated with the search word. However, the other search word is, for example, a search word for which one of the FAQs displayed on the first page for the other search word is selected.

  Specifically, for example, the generation unit 1005 acquires an access log with the same session number from the access log DB 230. Next, the generation unit 1005 sorts the acquired access log in chronological order based on the acquired date and time of the access log. Then, the generation unit 1005 determines whether or not there is an access log of type “inquiry” in the acquired access log.

  Here, if there is an access log of type “inquiry”, the generation unit 1005 determines that the inquiry FAQ is displayed on the first page based on the order of the inquiry FAQs identified from the access log, ie, the top Determine if there are N FAQs. Here, in the case of the FAQ displayed on the first page, teacher data (Yoiko) is generated, which represents the query FAQ specified from the access log of type “query” in association with the search word after morphological analysis. The morphologically-analyzed search word is one or more morphemes detected by morphologically analyzing a search word specified from an access log of type “query”.

  Further, the generation unit 1005 refers to the search list of the access log of the type “search”, and the FAQ which is not referred to when displayed for the search word, that is, the FAQ which did not receive the selection operation for the search word. Determine if there is. Here, when there is a FAQ that has not been inquired, the generation unit 1005 generates teacher data (badness) that sets the FAQ that has not been inquired as a non-inquiry FAQ. For example, a search word after morphological analysis of the teacher data (Rukoko) is obtained from a search word specified from an access log of type "query".

  Further, the generation unit 1005 refers to the search list of the access log of the type “search”, and determines whether there is a search word for which no FAQ has been searched. Here, when there is a search word for which even one FAQ has not been searched, the generation unit 1005 has, based on the acquired date and time of the access log, another search word that has been accepted following the search word. Determine if it is or not. However, the other search word is a search word for which one of the FAQs displayed on the first page for the other search word is inquired (selected). Here, when there is another search word, the generation unit 1005 generates teacher data (Magogo) in which the other search word is an alternative word. The morpheme-analyzed search word of the teacher data (music) is, for example, one or a plurality of morphemes detected by morphologically analyzing another search word.

  The generated teacher data (Yoshiko) is stored, for example, in the teacher data (Yokoko) DB 240 shown in FIG. The generated teacher data (raw data) is stored, for example, in the teacher data (bad data) DB 250 shown in FIG. The generated teacher data (music) is stored, for example, in the teacher data (data) DB 260 shown in FIG.

  An example of generation of teacher data (Youko), teacher data (Rusiko) and teacher data (Mago) will be described later with reference to FIGS. 11 to 14.

  In addition, when the teacher data (Yosukeko) and the teacher data (Rouiko) are not stored, the information processing apparatus 101 can not narrow down the FAQ using the first and second probabilities described above. Therefore, the information processing apparatus 101 may output the retrieved FAQ to the search word for a certain period of time from the start of operation. At this time, the information processing apparatus 101 may sort and output, for example, a plurality of searched FAQs in the order of the large number of accesses in the past (the number of inquiries). Describing in more detail, for example, the output control unit 1004 transmits the searched FAQ to the terminal 201, thereby displaying the FAQ for the received search word in the display area 910 of the FAQ screen 900. Do. This makes it possible to accumulate teacher data (Yosukeko) and teacher data (Uruiko).

  In addition, the determination unit 1003 may adjust the threshold value α based on the access log indicating the FAQ in which the selection operation on the search word has been received. Specifically, for example, based on the access log of type “inquiry” in the access log DB 230, the determination unit 1003 selects one of the top N FAQs for the search word, that is, any of the FAQs displayed on the first page. Calculate the probability that is selected.

  In the following description, the probability that one of the FAQs displayed on the first page is selected may be referred to as “first page hit ratio”.

  The calculation of the first page hit rate is performed, for example, at predetermined intervals. The predetermined period can be set arbitrarily, and is set, for example, to a period of several weeks to several months. At this time, the determination unit 1003 calculates the first page hit rate, for example, based on the access log for the latest predetermined period. Thereby, it is possible to calculate the first page hit rate in which the latest access tendency is reflected.

  Then, the determination unit 1003 adjusts the threshold value α based on the calculated time-series change of the first page hit rate. In more detail, for example, the determination unit 1003 lowers the threshold α when the first page hit rate tends to decrease. The reduction amount of the threshold value α can be arbitrarily set. For example, when the unit is [%], it is set to a value of about 1.

  An adjustment example of the threshold value α will be described later with reference to FIGS. 17 and 18.

  The determination unit 1006 determines the presence or absence of the FAQ corresponding to the search word. More specifically, for example, when no FAQ corresponding to the search word is searched from the FAQ master 220, the determination unit 1006 has no FAQ corresponding to the search word in the case of so-called 0 hits. judge.

  In addition, the determination unit 1006 may determine that there is no FAQ corresponding to the search word when the determination unit 1003 does not determine even one FAQ for the search word. Describing in more detail, for example, when the determination unit 1006 narrows down the FAQ using the first and second probabilities described above, if no FAQ remains, the FAQ corresponding to the search word It may be determined that there is no.

  When it is determined that there is no FAQ corresponding to the search word, the identification unit 1007 performs machine learning based on the storage content of the storage unit 1010 that stores teacher data (music), and is input following the search word. Identify other search words. In the following description, a search word determined by the determination unit 1006 to have no FAQ may be described as “0 hit search word”.

  Specifically, for example, the identifying unit 1007 performs machine learning with a naive Bayesian classifier with reference to the teacher data (music) DB 260, and 0 hit search words for each of the other search words (alternative words). And calculate a third probability to be input subsequently. Then, the specifying unit 1007 specifies an alternative word to be input following the zero hit search word based on the third probability calculated for each of the other search words.

  More specifically, for example, the specifying unit 1007 may specify another search word having the largest third probability as an alternative word. Further, the specifying unit 1007 may specify another search word whose third probability is the threshold value β or more as an alternative word. The threshold value β can be set arbitrarily. Further, the specifying unit 1007 may specify another search word having the third probability equal to or more than the threshold value β and the largest as an alternative word.

  The specific processing content for calculating the third probability will be described later with reference to FIG.

  The output control unit 1004 outputs the FAQ corresponding to the specified alternative word. Specifically, for example, the search unit 1002 first searches the FAQ corresponding to the alternative word. If a plurality of alternative words are specified, the search unit 1002 may search the FAQ for each of the plurality of alternative words.

  Next, the determination unit 1003 performs machine learning based on the storage content of the storage unit 1010, and determines one or more FAQs for the substitute word from among the plurality of searched FAQs. Then, the output control unit 1004 outputs the determined one or more FAQs to the alternative word. The output control unit 1004 may output the FAQ corresponding to the retrieved alternative word.

  Thus, even if there is no FAQ corresponding to the input search word, the search word, that is, an alternative word to be substituted for the 0 hit search word is estimated based on the operation history (access log) of the past user. Can output the FAQ.

(Example of teacher data generation)
Next, an example of generation of teacher data (わ る) and teacher data (ま い) will be described using FIGS. 11 to 14. Here, it is assumed that the top 5 FAQs are FAQs displayed on the first page (N = 5).

  FIG. 11 is an explanatory diagram (part 1) of a specific example of the access log. FIG. 12 is an explanatory view (part 1) of a specific example of teacher data. In FIG. 11, access logs 1101 to 1103 are a set of access logs having the same session number sorted in time series.

  In this case, the generation unit 1005 determines whether or not there is an access log of type “inquiry” in the access logs 1101 to 1103. Here, the access logs 1102 and 1103 are access logs of type “inquiry”. Therefore, the generation unit 1005 determines that there is an access log of type “inquiry”.

  Next, the generation unit 1005 determines whether the inquiry FAQ is the FAQ displayed on the first page based on the order of the access log 1102 of type “inquiry”. Here, the order of the access log 1102 is “1”, which is 5 or less. Therefore, the generation unit 1005 determines that the inquiry FAQ of the access log 1102 is the FAQ displayed on the first page.

  In this case, as illustrated in FIG. 12, the generation unit 1005 associates the inquiry FAQ “011” specified from the access log 1102 with the search word “magnify, print” after morpheme analysis, and indicates teacher data (Yosuke) 1201. Generate

  Similarly, the generation unit 1005 determines whether the inquiry FAQ is the FAQ displayed on the first page based on the order of the access log 1103 of type “inquiry”. Here, the order of the access log 1103 is “2”, which is 5 or less. Therefore, the generation unit 1005 determines that the inquiry FAQ in the access log 1103 is the FAQ displayed on the first page.

  In this case, as illustrated in FIG. 12, the generation unit 1005 associates the inquiry FAQ “012” specified from the access log 1103 with the search word “after expansion” and the search word “enlargement, print” in association with teacher data (Yosuke) 1202 Generate

  In addition, the generation unit 1005 refers to the search list of the access log 1101 of type “search”, and does not accept the FAQ which is not referred to when displayed for the search word, that is, the selection operation for the search word Determine if there is a FAQ. Here, the FAQ of the FAQ number "013" is not inquired.

  For this reason, the generation unit 1005 determines that there is a FAQ that has not been referred to. In this case, as shown in FIG. 12, the generation unit 1005 generates teacher data (bad) 1203 in which the uninquired FAQ having the FAQ number “013” is a non-inquired FAQ.

  Further, the generation unit 1005 refers to the search list of the access log of the type “search”, and determines whether there is a search word for which no FAQ has been searched. Here, there is no search word for which no FAQ has been searched. In this case, the generation unit 1005 does not generate teacher data.

  FIG. 13 is an explanatory view (part two) of a specific example of the access log. FIG. 14 is an explanatory view (part 2) of a specific example of teacher data. In FIG. 13, access logs 1301 to 1304 are a set of access logs having the same session number sorted in time series.

  In this case, the generation unit 1005 determines whether or not there is an access log of type “inquiry” in the access logs 1301 to 1304. Here, the access logs 1303 and 1304 are access logs of type “inquiry”. Therefore, the generation unit 1005 determines that there is an access log of type “inquiry”.

  Next, the generation unit 1005 determines whether the inquiry FAQ is the FAQ displayed on the first page based on the order of the access log 1303 of type “inquiry”. Here, the order of the access log 1303 is “1”, which is 5 or less. Therefore, the generation unit 1005 determines that the inquiry FAQ of the access log 1303 is the FAQ displayed on the first page.

  In this case, as illustrated in FIG. 14, the generation unit 1005 associates teacher data (Yosukeko) 1401 that represents the inquiry FAQ “011” specified from the access log 1303 in association with the search word “magnify, print” after morphological analysis. Generate

  Similarly, the generation unit 1005 determines whether the inquiry FAQ is the FAQ displayed on the first page based on the order of the access log 1304 of type “inquiry”. Here, the order of the access log 1304 is “2”, which is 5 or less. Therefore, the generation unit 1005 determines that the inquiry FAQ of the access log 1304 is the FAQ displayed on the first page.

  In this case, as illustrated in FIG. 14, the generation unit 1005 is teacher data (Yosukeko) 1402 that associates the inquiry FAQ “012” specified from the access log 1304 with the search word “magnify, print” after morphological analysis. Generate

  Further, the generation unit 1005 refers to the search list of the access log 1301 or 1302 of the type “search”, and accepts the FAQ not inquired when displayed for the search word, that is, the selection operation for the search word Determine if there is a FAQ that was not available. Here, the FAQ of the FAQ number "013" is not inquired.

  For this reason, the generation unit 1005 determines that there is a FAQ that has not been referred to. In this case, as illustrated in FIG. 14, the generation unit 1005 generates teacher data (badness) 1403 in which the uninquired FAQ having the FAQ number “013” is a non-inquired FAQ.

  In addition, the generation unit 1005 refers to the search list of the access log 1301 or 1302 of the type “search” to determine whether there is a search word for which no FAQ has been searched. Here, the search list of the access log 1301 is “−”, and the search word “large print” is a search word for which no FAQ has been searched.

  In this case, the generation unit 1005 determines whether there is another search word that has received an input following the search word “print large”. However, the other search word is a search word for which one of the FAQs displayed on the first page for the other search word is inquired (selected).

  Here, the search word “enlargement, print” is another search word that has received an input following the search word “large print”. Further, the search word “enlargement, print” is a search word for which the FAQ of the FAQ number “001, 002” displayed on the first page for the other search word is inquired (selected). Therefore, the generation unit 1005 determines that there is another search word.

  In this case, as illustrated in FIG. 14, the generation unit 1005 is teacher data in which the alternative word “enlargement, print” is associated with the search word “large print”, using another search word “enlargement, print” as an alternative word (Migo) 1404 is generated.

(Specific processing content to calculate the first, second and third probabilities)
Next, with reference to FIG. 15, specific processing contents for performing the machine learning by the naive Bayesian classifier to calculate the first, second, and third probabilities will be described.

  FIG. 15 is an explanatory diagram of an example of the storage content of the analysis DB 1500. Here, a calculation example will be described by taking the first probability as an example. First, the determination unit 1003 creates an analysis DB 1500 with reference to the teacher data (Yokoko) DB 240.

  The analysis DB 1500 sets “FAQ number” on the vertical axis and “keyword” on the horizontal axis, and counts the number of queries of each FAQ for each keyword. However, the FAQ number is the FAQ number of each of the plurality of FAQs searched by the search unit 1002. The keyword is a morpheme included in the search word after morpheme analysis of the teacher data (Yokoko) in the teacher data (Yokoko) DB 240.

  Here, the search word is “keyword 1 and keyword 2”. Further, the FAQ numbers of the plurality of FAQs corresponding to the search word “keyword 1, keyword 2” searched by the search unit 1002 are set to “1, 2, 3, 4, 5”. In addition, the morpheme contained in the search word after the morphological analysis of the teacher data (Yoshiko) in the teacher data (Yoshiko) DB 240 is set as “keyword1, keyword2, keyword3, keyword4, keyword5, keyword6”.

  Specifically, for example, the determination unit 1003 selects unselected teacher data (Yokoko) from the teacher data (Yokoko) DB 240. Here, it is assumed that the inquiry FAQ of the selected teacher data (Yokko) is “1”, and the search word after morphological analysis includes the keyword 1 and the keyword 2. In this case, the determination unit 1003 increments the number of queries of the keyword 1 of the FAQ number “1”. In addition, the determination unit 1003 increments the number of queries of the keyword 2 of the FAQ number “1”. The determination unit 1003 repeats the above-described series of processing until there is no unselected teacher data (Yokoko) not selected from the teacher data (Yokoko) DB 240.

  Next, the determination unit 1003 calculates a first probability for each of the plurality of FAQs searched with reference to the analysis DB 1500. Specifically, for example, the determination unit 1003 can calculate the first probability for each of the plurality of searched FAQs, for example, using the following equation (1).

    P (FAQ | search word) = P (FAQ) P (search word | FAQ) ... (1)

  The posterior probability P (FAQ | search word) indicates the probability that the corresponding FAQ is selected (inquired) when the FAQ is searched by designating the search word. That is, the posterior probability P (FAQ | search word) corresponds to the first probability. For example, P (FAQ 1 | accident) indicates the probability that FAQ 1 is selected (inquired) when the search word “accident” is specified and the FAQ is searched.

  When the search word is composed of a plurality of keywords, the posterior probability is indicated by the product of the respective probabilities. For example, if P (FAQ 1 | accident, car) specifies the search word "accident", "car" and searches the FAQ, the probability that FAQ 1 is selected is P (FAQ 1) P (accident | FAQ 1) × P (FAQ 1) Indicates that it is P (car | FAQ 1).

  The prior probability P (FAQ) indicates the probability (proportion) that the corresponding FAQ is selected. For example, when the total number of FAQs is 5000 and a search is performed using the search word "accident", P (FAQ 1) is "1/5000". The prior probability P (FAQ) is replaced with the total number of FAQs included in the search word in order to improve the accuracy in estimation. Assuming that the total number of FAQs is 5000 and the total number of FAQs that have been hit by the search word “accident” is 100, P (FAQ 1) is “1/100” when the search word “accident” is searched.

  The likelihood P (search word | FAQ) indicates the probability that a search word is created when the FAQ is determined. For example, P (accident | FAQ 1) indicates the probability that the search word “accident” is created from FAQ 1. If the number of accesses to the keyword that accessed FAQ 1 is 1000 and the search word “accident” is 10, P (accident | FAQ 1) becomes “10/100”.

  The first probabilities (P (FAQ | keywords 1 and 2)) of the respective FAQs 1 to 5 calculated based on the analysis DB 1500 shown in FIG. 15 are as follows. In addition, the part used as "0" is substituted to "1" (Laplace smoothing).

P (FAQ 1 | keywords 1 and 2)
= P (keyword 1 | FAQ 1) x P (FAQ 1) x P (keyword 2 | FAQ 1) x P (FAQ 1) = (1/21) (1/23) x (12/21) (12/96)
= 0.00015

P (FAQ 2 | keywords 1 and 2)
= P (keyword 1 | FAQ 2) x P (FAQ 2) x P (keyword 2 | FAQ 2) x P (FAQ 2) = (10/85) (10/23) x (1/85) (1/96)
= 0.00001

P (FAQ 3 | keywords 1 and 2)
= P (keyword 1 | FAQ 3) x P (FAQ 3) x P (keyword 2 | FAQ 3) x P (FAQ 3) = (2/7) (2/23) x (1/7) (1/96)
= 0.00018

P (FAQ 4 | keywords 1 and 2)
= P (keyword 1 | FAQ 4) x P (FAQ 4) x P (keyword 2 | FAQ 4) x P (FAQ 4) = (6/124) (6/23) x (77/124) (77/96)
= 0.00629

P (FAQ 5 | keywords 1 and 2)
= P (keyword 1 | FAQ 5) x P (FAQ 5) x P (keyword 2 | FAQ 5) x P (FAQ 5) = (4/44) (4/23) x (6/44) (6/96)
= 0.00013

  However, on a computer, it may not be possible to calculate if the number of digits becomes too small (overflow). Therefore, the determination unit 1003 converts the calculated values into logarithms to calculate the first probability. The first probability converted to logarithm is as follows. The closer to 0, the higher the probability.

P (FAQ 1 | keywords 1 and 2)
= Log (1/21) (1/23) + Log (12/21) (12/96)
=-3.83

P (FAQ 2 | keywords 1 and 2)
= Log (10/85) (10/23) + Log (1/85) (1/96)
=-5.20

P (FAQ 3 | keywords 1 and 2)
= Log (2/7) (2/23) + Log (1/7) (1/96)
= -4.43

P (FAQ 4 | keywords 1 and 2)
= Log (6/124) (6/23) + Log (77/124) (77/96)
=-2.2

P (FAQ 5 | keywords 1 and 2)
= Log (4/44) (4/23) + Log (6/44) (6/96)
=-3.87

  Next, the determination unit 1003 converts the first probability of each of the FAQs 1 to 5 logarithmically represented to 0 to 100 [%] using the soft max function. Specifically, for example, the determination unit 1003 acquires the maximum value C “−5.20” of the absolute value among the first probabilities of the FAQs 1 to 5 logarithmically represented. Then, the determination unit 1003 calculates an index of the value of (maximum value C-first probability) for each of the FAQs 1 to 5 as follows.

P (FAQ 1) = exp (-3. 83-(-5. 20)) = 3.93
P (FAQ 2) = exp (-5. 20-(-5. 20)) = 1.0
P (FAQ 3) = exp (-4. 43-(-5. 20)) = 2.16
P (FAQ 4) = exp (-2.20-(-5.20)) = 20.09
P (FAQ 5) = exp (-3. 87-(-5. 20)) = 3.42.

  Next, the determination unit 1003 calculates the sum value S of the exponents of the values (maximum value C-first probability) of each of the FAQs 1 to 5. Here, the total value S of the indices is "30.06".

  Next, the determination unit 1003 calculates a value obtained by dividing the exponent of the value of (maximum value C-first probability) by the total value S of the exponents for each of the FAQs 1 to 5 as follows.

P (FAQ 1) = 3.93 / 3.0.06 = 0.13
P (FAQ 2) = 1.0 / 3.06 = 0.03
P (FAQ 3) = 2.16 / 3.06 = 0.07
P (FAQ 4) = 20.09 / 03.06 = 0.67
P (FAQ 5) = 3.42 / 3 0.06 = 0.11

  Next, the determination unit 1003 converts the first probability into a notation of 0 to 100 [%] by multiplying the calculated value by 100 for each of the FAQs 1 to 5 as follows.

P (FAQ 1) = 13 [%]
P (FAQ 2) = 3 [%]
P (FAQ 3) = 7 [%]
P (FAQ 4) = 67 [%]
P (FAQ 5) = 11 [%]

  Thus, for each of the FAQs 1 to 5, when each of the FAQs 1 to 5 is displayed for the search word "keyword 1, keyword 2", the first probability that each of the FAQs 1 to 5 is selected (inquired) is It can be calculated. The first probability is “0 [%]” for FAQs that have not been selected (inquired) even once.

  Further, although the first probability is described as an example here, the second and third probabilities can be similarly obtained.

  For example, in the case of the second probability, the determination unit 1003 creates the analysis DB 1500 with reference to the teacher data (bad) DB 250. In this case, the analysis DB 1500 sets “FAQ number” on the vertical axis and “keyword” on the horizontal axis, and counts the number of non-inquiries for which each FAQ has not been queried for each keyword. However, the FAQ number is the FAQ number of each of the plurality of FAQs searched by the search unit 1002. The keyword is a morpheme included in the search word after morpheme analysis of the teacher data (Rusiko) in the teacher data (Rumiko) DB 250.

  Specifically, for example, the determination unit 1003 selects unselected teacher data (roiko) from the teacher data (roko) DB 250. Here, it is assumed that the non-inquiring FAQ of the selected teacher data (bad box) is “1”, and the search word after morphological analysis includes the keyword 1 and the keyword 2. In this case, the determination unit 1003 increments the number of non-inquiries of the keyword 1 of the FAQ number “1”. In addition, the determination unit 1003 increments the number of non-inquiries of the keyword 2 of the FAQ number “1”. The determination unit 1003 repeats the above-described series of processing until there is no unselected teacher data (Yokoko) not selected from the teacher data (Yokoko) DB 240.

  Thereby, it is possible to create an analysis DB 1500 in which the number of non-inquiries of each FAQ is stored for each keyword. Then, the determination unit 1003 calculates, for example, the second probability for each of the plurality of FAQs searched using the above equation (1) with reference to the analysis DB 1500. However, the posterior probability P (FAQ | search word) indicates the probability that the corresponding FAQ is not selected (inquired) when the FAQ is searched by designating the search word. The prior probability P (FAQ) indicates the probability (proportion) that the corresponding FAQ is not selected. The likelihood P (search word | FAQ) indicates the probability that a search word is created when the FAQ is determined. In addition, since the specific calculation procedure of subsequent 2nd probability is the same as 1st probability, detailed description is abbreviate | omitted.

  Also, in the case of the third probability, the identifying unit 1007 creates the analysis DB 1500 with reference to the teacher data (music) DB 260. In this case, the analysis DB 1500 sets “alternate word” on the vertical axis and “morpheme” on the horizontal axis, and counts the number of times each alternative word is input for each morpheme. However, the alternative word is an alternative word of teacher data (mygo) in the teacher data (mygo) DB 260. That is, the substitute word corresponds to another search word input following the zero hit search word. A morpheme is a morpheme detected by morpheme analysis of a zero hit search word. If no morpheme is detected from the zero hit search word, for example, the zero hit search word itself may be taken as the morpheme on the horizontal axis.

  Specifically, for example, the identifying unit 1007 selects unselected teacher data (music) from the teacher data (music) DB 260. Then, when the search word of the selected teacher data (Magogo) includes any morpheme in the analysis DB 1500, the identifying unit 1007 determines the number of times of input of the substitute word of the selected teacher data (Magogo) for that morpheme. Increment. The identifying unit 1007 repeats the above-described series of processing until there is no unselected teacher data (mago) not selected from the teacher data (mago) DB 260.

  Thus, it is possible to create an analysis DB 1500 in which the number of times each alternative word has been input is stored for each morpheme. Then, the identifying unit 1007 calculates, for example, the third probability for each of the alternative words in the analysis DB 1500 using the above equation (1) with reference to the analysis DB 1500.

  However, the posterior probability is P (alternative word | search word). The prior probability is P (alternate word). The likelihood is P (alternative word). The posterior probability P (Alternative Word | Search Word) indicates the probability that an alternative word is input following the search word. The prior probability P (alternate word) indicates the probability (proportion) that the relevant alternative word is present. The likelihood P (search word | alternate word) indicates the probability that a search word is created when an alternative word is determined. In addition, since the specific calculation procedure of subsequent 3rd probability is the same as 1st probability, detailed description is abbreviate | omitted.

(Example of decision of FAQ)
Next, a determination example of the FAQ for the search word will be described with reference to FIG.

  FIG. 16 is an explanatory diagram of an example of determination of the FAQ. In FIG. 16, the first table 1601 stores the first probability calculated for each of the plurality of FAQs searched by the search unit 1002. The second table 1602 stores the second probability calculated for each of the plurality of FAQs searched by the search unit 1002.

  Here, it is assumed that ten FAQs with FAQ numbers “001 to 010” are searched as FAQs corresponding to search words. Further, the threshold value α is set to “α = 80”. Also, the lower the FAQ number, the higher the priority when retrieved using the existing search algorithm. For example, in the existing search algorithm, the higher the number of accesses in the past (the number of queries), the higher the priority is set.

  In this case, the determining unit 1003 specifies, for example, the FAQ whose second probability is equal to or higher than the threshold α with reference to the second table 1602. Here, the FAQ with the FAQ number "003, 005, 006, 008" is identified. Next, the determination unit 1003 deletes the FAQ record of the identified FAQ number “003, 005, 006, 008” from the first table 1601.

  Then, the determination unit 1003 refers to the first table 1601 and sorts the remaining FAQs sorted in the descending order of the first probability, that is, the FAQs of the FAQ numbers “001, 002, 007, 004, 009, 010”. Decide on one or more FAQs for the search word.

  In this case, the output control unit 1004 transmits the FAQ of the determined FAQ number “001, 002, 007, 004, 009, 010” to the terminal 201. Here, in the FAQ screen 900, when only the top 5 items are displayed on the first page (N = 5), the FAQs with the FAQ number “001, 002, 007, 004, 009” are the first page in this order. As it is displayed, the FAQ with the FAQ number "010" is displayed on the second page.

  For example, as a result of the FAQ being narrowed down using the first and second probabilities, the FAQ number "003, 005, 006, 008" as compared to the case where the FAQ is displayed in descending order of the number of accesses (number of queries) in the past. The FAQ is excluded, and the FAQ with the FAQ number "007, 009, 010" is displayed on the first page.

(Example of adjustment of threshold α)
Next, an adjustment example of the threshold value α will be described with reference to FIGS. 17 and 18. Here, it is assumed that the top 5 FAQs are FAQs displayed on the first page (N = 5).

  FIG. 17 is an explanatory view (part 3) of a specific example of the access log. In FIG. 17, access logs 1701 to 1705 are a set of access logs having the same session number sorted in time series. Hereinafter, the case of calculating the first page hit rate from the access logs 1701 to 1705 will be described.

  Specifically, for example, the determination unit 1003 calculates the first page hit rate from the access logs 1701 to 1705 using the following formula (2). Where X is the number of times the inquired FAQ is present on the first page of the search results. Y is the number of inquiries for the FAQ. Also, the unit of the first page hit rate is [%].

    First page hit rate = X / Y × 100 (2)

  More specifically, the determination unit 1003 calculates the number of times X in which the inquired FAQ is present on the first page of the search result with reference to the access log 1702-1705 of the type “inquiry”. Here, the order of the access logs 1702 and 1703 is “1, 2” and is 5 or less. On the other hand, the order of the access logs 1704 and 1705 is “40, 45” and is greater than 5. Therefore, the number X is “X = 2”.

  The determination unit 1003 also calculates the number of times of inquiry Y of the FAQ with reference to the access logs 1702 to 1705 of the type “inquiry”. Here, the number of times of inquiry Y is the number of access logs 1702-1705 of type "inquiry", and is "Y = 4". For this reason, the first page hit rate is “50 [%] (= 2/4 × 100)”.

  The determination unit 1003 calculates the first page hit rate every predetermined period, for example, every several weeks to several months. Next, the determination unit 1003 performs regression analysis of the transition of the first page hit ratio calculated for each predetermined period, and calculates the slope of a straight line representing the time-series change of the first page hit ratio. Then, the determination unit 1003 adjusts the threshold value α according to the calculated inclination.

  FIG. 18 is an explanatory diagram of an example of time-series change of the first page hit rate. In FIG. 18, straight lines 1801 and 1802 are straight lines representing time-series changes in the first page hit rate. A straight line 1801 indicates that the slope is greater than 0 and the hit ratio for the first page tends to increase. On the other hand, a straight line 1802 indicates that the slope is smaller than 0 and the hit ratio for the first page tends to decrease.

  For example, when the slope of the straight line representing the time-series change of the first page hit rate is larger than 0, the determining unit 1003 does not adjust the threshold value α because the first page hit rate tends to increase. When the slope of the straight line representing the time-series change of the first page hit rate is 0, the determination unit 1003 does not adjust the threshold value α because there is no change in the first page hit rate.

  On the other hand, when the slope of the straight line representing the time-series change of the first page hit ratio is smaller than 0, the determination unit 1003 adjusts the threshold value α because the first page hit ratio tends to decrease. Specifically, for example, when the slope of the straight line representing the time-series change of the first page hit rate is smaller than 0, the determination unit 1003 lowers the threshold α by 1 [%].

  For example, when the threshold α before adjustment is “80 [%]”, the determination unit 1003 changes the threshold α to “79 [%]”. If the hit ratio of the first page is in a downward trend, it is assumed that the FAQ for the first page is less likely to be referred to, and learning of teacher data (scoring) proceeds. For this reason, in order to improve the response of the teacher data (bad texture), the threshold value α is lowered to adjust the FAQ in a direction to cut more.

  The determination unit 1003 may change the decrease of the threshold α in accordance with the magnitude of the slope of the straight line that represents the time-series change of the first page hit rate. For example, if the slope of the straight line is small to such an extent that it can be determined that the first page hit rate has dropped significantly, the determination unit 1003 may change the decrease of the threshold α to about 5%. Good.

(Estimate of morpheme)
Next, processing for estimating a morpheme will be described using FIG.

  FIG. 19 is an explanatory diagram of an example of a method of estimating morphemes. When it is determined by the determination unit 1006 that there is no FAQ corresponding to the search word, the identification unit 1007 performs machine learning based on the storage content of the teacher data (music) DB 260, and inputs the zero hit search word following Identify the alternative word to be However, alternative words can not be identified until teacher data is accumulated.

  For example, in the case where a tedious search word is input due to a punch error or the like, morphemes may not be detected from the search word, resulting in zero hits. Further, in this case, there is a high possibility that the teacher data (music) corresponding to the search word is not stored. That is, in the case where a search word is corrupted due to a punch error or the like, an alternative word can not be identified, resulting in zero hits.

  Therefore, when the alternative word is not identified, the information processing apparatus 101 may estimate the morpheme corresponding to the search word and search the FAQ corresponding to the estimated morpheme.

  Specifically, for example, first, the generation unit 1005 generates teacher data (morpheme). More specifically, for example, the generation unit 1005 extracts a morpheme from the search word after morpheme analysis of the teacher data (Yoshiko) in the teacher data (Yoshiko) DB 240. Then, the generation unit 1005 generates teacher data (morpheme) in which the extracted morpheme is associated with each character included in the morpheme.

  As an example, when the morpheme is "FILE", the morpheme "FILE" and the respective elements (characters) "F", "I", "L", "E" included in the morpheme "FILE" correspond to each other The attached teacher data (morpheme) 1901 is generated. The generated teacher data (morpheme) is stored in the teacher data (morpheme) DB 1900.

  In addition, for example, the generation unit 1005 extracts a morpheme from the search word after morpheme analysis of the teacher data (rokko) in the teacher data (rokko) DB 250, and extracts the extracted morpheme and each character included in the morpheme You may generate the teacher data (morpheme) which matched the. In addition, the generation unit 1005 extracts morphemes from, for example, substitute words or search words of teacher data (maigo) in the teacher data (maigo) DB 260, and associates the extracted morphemes with the respective characters included in the morphemes. You may generate attached teacher data (morpheme).

  Next, when the morpheme is not detected from the search word, the identifying unit 1007 performs machine learning based on the storage content of the teacher data (morpheme) DB 1900 to identify the morpheme corresponding to the search word. However, even if no morpheme is detected from the search word, the FAQ may be searched from the FAQ master 220.

  That is, even if no morpheme is detected from the search word, there may be a case where the FAQ corresponding to the search word can be output to the terminal 201. For this reason, the identifying unit 1007 may identify the morpheme corresponding to the search word, for example, when it is determined that the morpheme is not detected from the search word and there is no FAQ corresponding to the search word. .

  More specifically, for example, the identifying unit 1007 performs machine learning with a naive Bayesian classifier with reference to the teacher data (morpheme) DB 1900, and calculates a fourth probability for each of the plurality of morphemes. The fourth probability indicates the degree of similarity to the search word for which the morpheme was not detected.

  The fourth probability can also be obtained in the same manner as the first probability. First, the identifying unit 1007 creates the analysis DB 1500 with reference to the teacher data (morpheme) DB 1900. In this case, the analysis DB 1500 sets “morpheme” on the vertical axis and “character” on the horizontal axis, and counts the number of appearances for each character. However, the morpheme is a morpheme of teacher data (morpheme) in the teacher data (morpheme) DB 1900. The characters are characters included in the morpheme of the teacher data (morpheme) in the teacher data (morpheme) DB 1900.

  Specifically, for example, the specifying unit 1007 selects unselected teacher data (morpheme) from the teacher data (morpheme) DB 1900. Here, it is assumed that the morpheme of the selected teacher data (morpheme) is "FILE", and the elements are "F", "I", "L", "E". In this case, the identifying unit 1007 increments the number of occurrences of each element “F”, “I”, “L”, and “E” of the morpheme “FILE” in the analysis DB 1500. The identifying unit 1007 repeats the above-described series of processes, for example, until there is no unselected teacher data (morpheme) not selected from the teacher data (morpheme) DB 1900.

  Thus, it is possible to create an analysis DB 1500 in which the number of occurrences of each element is stored for each morpheme. Then, the determination unit 1003 calculates the fourth probability for each of the plurality of morphemes using, for example, the above-mentioned equation (1) with reference to the analysis DB 1500. However, the posterior probability is P (morpheme character). The prior probability is P (morpheme). The likelihood is P (morpheme). The posterior probability P (morpheme | character) indicates the probability that a morpheme including the corresponding character is input. The prior probability P (morpheme) indicates the proportion of the corresponding character in all elements included in the morpheme (vertical axis). The likelihood P (morpheme) indicates the probability that the corresponding character is included when the morpheme is determined. In addition, since the specific calculation procedure of subsequent 4th probability is the same as 1st probability, detailed description is abbreviate | omitted.

  Then, the identifying unit 1007 identifies the morpheme corresponding to the search word based on the fourth probability calculated for each of the plurality of morphemes in the analysis DB 1500. Specifically, for example, the identifying unit 1007 identifies the morpheme with the largest fourth probability as the morpheme corresponding to the form word. Thereby, it is possible to estimate the morpheme corresponding to the search word for which the morpheme was not detected by the morpheme analysis.

  In this case, the specifying unit 1007 performs machine learning based on the storage content of the storage unit 1010 that stores, for example, teacher data (music), and specifies an alternative word that is input following the identified morpheme. Then, the search unit 1002 searches the FAQ corresponding to the specified alternative word.

  This makes it possible to prevent zero hits even when a search word that is not correct due to a punch error or the like is input. Further, in the above description, the specifying unit 1007 specifies the alternative word input following the specified morpheme, and the searching unit 1002 searches for the FAQ corresponding to the specified alternative word, It is not limited to this. For example, the search unit 1002 may search the FAQ corresponding to the identified morpheme.

  In addition, the data structure of teacher data (Yosukeko) and teacher data (roiko) can not be learned if a morpheme is not detected from the search word because the FAQ number and the search word after morpheme analysis, the first and the first The probability of 2 can not be calculated. That is, it is not possible to narrow down the FAQ using the first and second probabilities. Therefore, even in such a case, the information processing apparatus 101 estimates the morpheme corresponding to the search word and re-searches the FAQ so that the FAQ can be narrowed down using the first and second probabilities. May be

(Procedure of Teacher Data Generation Processing of Information Processing Apparatus 101)
Next, a teaching data generation processing procedure of the information processing apparatus 101 will be described using FIGS. 20 and 21. FIG. The teacher data generation process of the information processing apparatus 101 may be performed periodically, for example, every 24 hours, or may be performed at a predetermined timing designated by the administrator of the response output system 200 or the like. Good.

  FIG. 20 and FIG. 21 are flowcharts showing an example of the teaching data generation processing procedure of the information processing apparatus 101. In the flowchart of FIG. 20, first, the information processing apparatus 101 selects an unselected session number not selected from the access log DB 230 (step S2001).

  Next, the information processing apparatus 101 acquires an access log of the selected session number from the access log DB 230 (step S2002). Then, based on the acquired date and time of the access log, the information processing apparatus 101 sorts the acquired access log in time series (step S2003).

  Next, the information processing apparatus 101 determines whether the acquired access log includes an access log of type “inquiry” (step S2004). Here, when there is no access log of type “inquiry” (step S2004: No), the information processing apparatus 101 proceeds to step S2101 shown in FIG.

  On the other hand, when there is an access log of type “inquiry” (step S2004: Yes), the information processing apparatus 101 determines whether the inquiry FAQ specified from the access log is the FAQ displayed on the first page. It judges (step S2005). Here, when it is not the FAQ displayed on the first page (step S2005: No), the information processing apparatus 101 proceeds to step S2101 shown in FIG.

  On the other hand, in the case of the FAQ displayed on the first page (step S2005: Yes), the information processing apparatus 101 displays the query FAQ specified from the access log of type “query” in association with the search word after morphological analysis. Teacher data (Yokko) is generated (step S2006). Then, the information processing apparatus 101 registers the generated teacher data (Yokko) in the teacher data (Yokko) DB 240 (step S2007), and shifts to step S2101 shown in FIG.

  In the flowchart of FIG. 21, first, the information processing apparatus 101 determines whether there is a search word of 0 hits with reference to the search list of the access log of type “search” (step S2101). Here, when there is no search word of 0 hits (step S2101: No), the information processing apparatus 101 proceeds to step S2105.

  On the other hand, when there is a search word of 0 hits (step S2101: Yes), the information processing apparatus 101 receives an input following the search word of 0 hits based on the acquired date and time of the access log. It is determined whether there is a search word (step S2102). However, the other search word is a search word for which one of the FAQs displayed on the first page for the other search word is inquired (selected).

  Here, when there is no other search word (step S2102: No), the information processing apparatus 101 proceeds to step S2105. On the other hand, when there is another search word (step S2102: YES), the information processing apparatus 101 generates teacher data (music) in which the other search word is an alternative word (step S2103).

  Then, the information processing apparatus 101 registers the generated teacher data (music) in the teacher data (data) DB 260 (step S2104). Next, the information processing apparatus 101 refers to the search list of the access log of the type “search” to determine whether there is a FAQ that is not referred to when displayed for the search word (step S2105). ).

  Here, when there is no FAQ that has not been referred to (step S2105: No), the information processing apparatus 101 proceeds to step S2108. On the other hand, if there is a FAQ that has not been inquired (step S2105: YES), the information processing apparatus 101 generates teacher data (badness) in which the FAQ that has not been inquired is a non-inquiry FAQ (step S2106).

  Then, the information processing apparatus 101 registers the generated teacher data (bad data) in the teacher data (bad data) DB 250 (step S2107). Next, the information processing apparatus 101 determines whether there is an unselected session number not selected from the access log DB 230 (step S2108).

  Here, if there is an unselected session number (step S2108: YES), the information processing apparatus 101 returns to step S2001 shown in FIG. On the other hand, when there is no unselected session number (step S2108: No), the information processing apparatus 101 ends the series of processing according to this flowchart.

  As a result, it is possible to generate teacher data (Yoiko), teacher data (Rusiko) and teacher data (Magogo) from the past user's operation history (access log). The stored contents of the access log DB 230 are reset each time, for example, when the execution of the teacher data generation process of the information processing apparatus 101 is completed.

(Reply output processing procedure of the information processing apparatus 101)
Next, the answer output processing procedure of the information processing apparatus 101 will be described using FIG. 22 and FIG.

  22 and 23 are flowcharts illustrating an example of the response output processing procedure of the information processing apparatus 101. In the flowchart of FIG. 22, first, the information processing apparatus 101 determines whether an input of a search word has been received (step S2201). Here, the information processing apparatus 101 waits to receive an input of a search word (step S2201: No).

  Then, when the information processing apparatus 101 receives the input of the search word (step S2201: Yes), the information processing apparatus 101 morphologically analyzes the search word (step S2202). Next, the information processing apparatus 101 searches the FAQ master 220 for the FAQ based on the result of morphological analysis of the search word (step S2203).

  Next, the information processing apparatus 101 performs machine learning with a naive Bayesian classifier with reference to the teacher data (Yokko) DB 240, and calculates a first probability for each of the plurality of searched FAQs (step S2204). ). Next, the information processing apparatus 101 performs machine learning using a naive Bayesian classifier with reference to the teacher data (bad data) DB 250, and calculates a second probability for each of the plurality of searched FAQs (step S2205).

  Then, the information processing apparatus 101 selects one or more FAQs for the search word from among the plurality of FAQs searched based on the first probability and the second probability calculated for each of the plurality of searched FAQs. Are determined (step S2206).

  Next, the information processing apparatus 101 determines whether there is a FAQ corresponding to the search word (step S2207). Specifically, for example, when the information processing apparatus 101 determines that no FAQ has been searched in step S 2203 or no FAQ has been determined in step S 2206, the information processing apparatus 101 corresponds to the FAQ corresponding to the search word. Determine that there is no

  Here, when there is a FAQ corresponding to the search word (step S2207: Yes), the information processing apparatus 101 outputs one or more FAQs for the determined search word (step S2208), and a series of processes according to this flowchart Finish. On the other hand, when there is no FAQ corresponding to the search word (step S2207: No), the information processing apparatus 101 proceeds to step S2301 shown in FIG.

  In the flowchart of FIG. 23, first, the information processing apparatus 101 performs machine learning by the naive Bayesian classifier with reference to the teacher data (data) DB 260, and inputs a search word following each of a plurality of alternative words. A third probability to be calculated is calculated (step S2301).

  Then, the information processing apparatus 101 identifies an alternative word to be input following the search word, based on the third probability calculated for each of the plurality of alternative words (step S2302). Next, the information processing apparatus 101 determines whether an alternative word has been identified (step S2303). For example, in step S2202 shown in FIG. 22, when a morpheme is not detected from the search word, an alternative word may not be identified.

  Here, if the alternative word is specified (step S2303: YES), the information processing apparatus 101 returns to step S2202 shown in FIG. In this case, the information processing apparatus 101 performs morphological analysis of the identified alternative word.

  On the other hand, when the alternative word is not specified (step S2303: No), the information processing apparatus 101 refers to the teacher data (morpheme) DB 1900, performs machine learning with the naive Bayesian classifier, and performs each of the plurality of morphemes. The fourth probability is calculated (step S2304). The fourth probability indicates the degree of similarity to the search word.

  Then, the information processing apparatus 101 identifies the morpheme corresponding to the search word based on the fourth probability calculated for each of the plurality of morphemes (step S2305), and returns to step S2203 illustrated in FIG. In this case, the information processing apparatus 101 searches the FAQ master 220 for the FAQ based on the identified morpheme.

  If no FAQ is found in step S2203, the information processing apparatus 101 may shift to step S2304. Further, the transition from step S2305 to step S2203 is not performed twice or more. Therefore, if there is no FAQ corresponding to the search word in step S2207 after transition (step S2207: No), the information processing apparatus 101 outputs, for example, information indicating that the search result is 0.

  This makes it possible to optimize the FAQ displayed for the search word based on the past operation history (access log) of the user. In addition, even if there is no FAQ corresponding to the search word, it is possible to estimate the search word, that is, an alternative word to be substituted for the zero hit search word based on the user's operation history in the past, and output the FAQ .

  As described above, according to the information processing apparatus 101 according to the embodiment, when there is no FAQ corresponding to the received search word, machine learning based on the storage content of the storage unit 1010 that stores teacher data is used. By doing this, it is possible to specify an alternative word to be input following the search word, and to output an FAQ corresponding to the specified alternative word. The teacher data (data) is information in which another search word, which is subsequently accepted when there is no FAQ corresponding to the search word, is associated with the search word. The other search word is, for example, a search word in which one of the FAQs displayed on the first page (upper predetermined number N) is selected with respect to the other search word.

  Thus, even if there is no FAQ corresponding to the input search word, the search word, that is, an alternative word to be substituted for the 0 hit search word is estimated based on the operation history (access log) of the past user. Can output the FAQ. Therefore, under the circumstances where the review of the FAQ can not keep up with the name change of the product or service, etc., or when there is a typo or conversion error at the time of search, display a proper FAQ that leads to the solution of the problem. Can.

  In addition, according to the information processing apparatus 101, machine learning is performed by the naive Bayesian classifier with reference to the storage unit 1010, and the other search words (alternative words) specified from the teacher data (music) are compared with each other. The probability of 3 can be calculated. The third probability is the probability of being input following the search word (0 hit search word). Then, according to the information processing apparatus 101, based on the third probability calculated for each of the other search words, it is possible to specify an alternative word to be input following the search word (zero hit search word). .

  This makes it possible to accurately estimate alternative words that can obtain FAQs that lead to problem solving. In addition, by using machine learning with a naive Bayesian classifier, it is possible to control the estimation accuracy of alternative words by the quality and amount of teacher data.

  Further, according to the information processing apparatus 101, when the alternative word is not identified, machine learning based on the storage content of the teacher data (morpheme) DB 1900 is performed to identify the morpheme corresponding to the search word. The corresponding FAQ can be output.

  In this way, even if a zero hit search word is input until the teacher data (mago) is stored, the morpheme associated with the zero hit search word is estimated, and the FAQ corresponding to the morpheme is calculated. , Can be displayed as a FAQ corresponding to the search word. For this reason, it is possible to display the FAQ related to the search word even when the search word is damaged due to a punch error or the like until the teacher data (music) is accumulated.

  From the above, according to the information processing apparatus 101, it is possible to improve the primary response rate and the customer satisfaction in the FAQ system and the FAQ site. In addition, it is possible to reduce the operation cost of the FAQ system and the FAQ site, because appropriate search results can be obtained without reviewing the FAQ.

  The response output method described in the present embodiment can be realized by executing a prepared program on a computer such as a personal computer or a workstation. This answer output program is a computer readable recording medium such as a hard disk, a flexible disk, a CD (Compact Disc) -ROM, a MO (Magneto-Optical disk), a DVD (Digital Versatile Disk), and a USB (Universal Serial Bus) memory. And read out from the recording medium by a computer. Also, the response output program may be distributed via a network such as the Internet.

  The following appendices will be further disclosed regarding the embodiment described above.

(Supplementary Note 1) Accepting the input of the first question,
In the case where there is no answer candidate corresponding to the accepted first question, it is another question which subsequently receives an input when there is no answer candidate corresponding to the question, and is displayed for the other question A machine learning is performed based on the memory content of the storage unit that stores the other question in which any answer candidate is selected in association with the question, and the second question is input following the first question Identify the questions of
Outputting an answer candidate corresponding to the identified second question;
An answer output program characterized by causing a computer to execute processing.

(Supplementary Note 2) The process for specifying
With reference to the storage unit, machine learning is performed by a naive Bayesian classifier to calculate, for each of the other questions, the probability of being input following the first question.
The answer output program according to appendix 1, wherein a second question input following the first question is specified based on the probability calculated for each of the other questions.

(Supplementary Note 3) The other question is another question that is subsequently input when there is no answer candidate corresponding to the question, and one of the upper predetermined number of answer candidates for the other question is selected The answer output program according to appendix 1 or 2, characterized in that it is a question that has been asked.

(Supplementary Note 4) If the second question is not identified, storage of a second storage unit that associates and stores a morpheme detected by morphologically analyzing the question and each element included in the morpheme Perform content-based machine learning to identify morphemes corresponding to the first question;
Outputting an answer candidate corresponding to the identified morpheme;
The answer output program according to any one of appendices 1 to 3, wherein the computer is caused to execute processing.

(Supplementary Note 5) The answer output program according to any one of supplementary notes 1 to 4, wherein the answer candidate is a frequently asked question (FAQ).

(Supplementary Note 6)
The question having the highest probability among the other questions is specified as a second question to be input following the first question, as described in any one of appendices 1 to 5 Answer output program.

(Supplementary Note 7) Accepting the input of the first question,
In the case where there is no answer candidate corresponding to the accepted first question, it is another question which subsequently receives an input when there is no answer candidate corresponding to the question, and is displayed for the other question A machine learning is performed based on the memory content of the storage unit that stores the other question in which any answer candidate is selected in association with the question, and the second question is input following the first question Identify the questions of
Outputting an answer candidate corresponding to the identified second question;
An answer output method characterized in that a computer executes a process.

(Supplementary Note 8) Accepting the input of the first question,
In the case where there is no answer candidate corresponding to the accepted first question, it is another question which subsequently receives an input when there is no answer candidate corresponding to the question, and is displayed for the other question A machine learning is performed based on the memory content of the storage unit that stores the other question in which any answer candidate is selected in association with the question, and the second question is input following the first question Identify the questions of
Outputting an answer candidate corresponding to the identified second question;
An information processing apparatus comprising a control unit.

101 information processing apparatus 110, 1010 storage unit 200 reply output system 201 terminal 210 network 220 FAQ master 230 access log DB
240 teacher data (Yoshiko) DB
250 teacher data (bad boy) DB
260 Teacher Data (Mygo) DB
300 bus 301 CPU
302 Memory 303 I / F
304 disk drive 305 disk 900 FAQ screen 1001 reception unit 1002 search unit 1003 determination unit 1004 output control unit 1005 generation unit 1005 determination unit 1007 identification unit 1101, 1102, 1103, 1301, 1302, 1303, 1304, 1701, 1702, 1703 1704, 1705 Access Log 1500 Analysis DB
1900 Teacher data (morpheme) DB

Claims (6)

Accept input for the first question,
In the case where there is no answer candidate corresponding to the accepted first question, it is another question which subsequently receives an input when there is no answer candidate corresponding to the question, and is displayed for the other question A machine learning is performed based on the memory content of the storage unit that stores the other question in which any answer candidate is selected in association with the question, and the second question is input following the first question Identify the questions of
Outputting an answer candidate corresponding to the identified second question;
An answer output program characterized by causing a computer to execute processing. The process for specifying is
With reference to the storage unit, machine learning is performed by a naive Bayesian classifier to calculate, for each of the other questions, the probability of being input following the first question.
The answer output program according to claim 1, wherein a second question input following the first question is specified based on the probability calculated for each of the other questions.   The other question is another question that is subsequently input when there is no answer candidate corresponding to the question, and one of the upper predetermined number of answer candidates for the other question is a selected question The answer output program according to claim 1 or 2, characterized in that: If the second question is not identified, a machine based on the memory content of the second storage unit that associates and stores a morpheme detected by morphologically analyzing the question and each element included in the morpheme Learning to identify morphemes corresponding to the first question,
Outputting an answer candidate corresponding to the identified morpheme;
The answer output program according to any one of claims 1 to 3, wherein the computer is made to execute processing. Accept input for the first question,
In the case where there is no answer candidate corresponding to the accepted first question, it is another question which subsequently receives an input when there is no answer candidate corresponding to the question, and is displayed for the other question A machine learning is performed based on the memory content of the storage unit that stores the other question in which any answer candidate is selected in association with the question, and the second question is input following the first question Identify the questions of
Outputting an answer candidate corresponding to the identified second question;
An answer output method characterized in that a computer executes a process. Accept input for the first question,
In the case where there is no answer candidate corresponding to the accepted first question, it is another question which subsequently receives an input when there is no answer candidate corresponding to the question, and is displayed for the other question A machine learning is performed based on the memory content of the storage unit that stores the other question in which any answer candidate is selected in association with the question, and the second question is input following the first question Identify the questions of
Outputting an answer candidate corresponding to the identified second question;
An information processing apparatus comprising a control unit.

Images