JP7126865B2 - Interactive business support system - Google Patents

Description

TECHNICAL FIELD The present invention relates to a technique for dealing with customers, and more particularly to a technique effectively applied to an interactive business support system for supporting the business of answering inquiries from customers.

For example, a business that sells products or provides services to customers such as consumers has a business of answering questions and inquiries from customers. In general, help desks and call centers accept and respond to inquiries by filling in forms on e-mails, websites, etc., and respond to inquiries by telephone and respond immediately. In recent years, VPAs (Virtual Personal Assistants), which accept inquiries by voice, chat, etc. via dedicated devices and applications and use AI (Artificial Intelligence) to immediately provide appropriate answers and responses, have become popular. I'm doing it.

As a mechanism for analyzing the contents of inquiries from customers in natural language, for example, Japanese Patent Application Laid-Open No. 2011-134291 (Patent Document 1) discloses a discourse structure analysis system that takes discourse data as input and outputs discourse semantics. is described. Here, there is a flow analysis rule consisting of a flow expression for determining flow information of each statement and a flow definition specifying a flow to be set for a statement having an expression matching the flow expression. Then, each statement is matched with each flow expression in the flow analysis rule, the flow specified by the corresponding flow definition in the matched flow expression is set, and the flow consists of correspondence between each statement and the flow. It has a flow analysis unit that outputs information, and outputs flow information as discourse semantics.

JP 2011-134291 A

For example, when an operator or communicator at a help desk, call center, etc. (hereinafter collectively referred to as "operator") accepts and responds to telephone inquiries from customers, the contents of the customer's inquiries are not complete from the beginning. It is rare that the text is comprehensible, and in many cases the content is insufficient, such as missing some information. In this case, the operator needs to ask additional questions to obtain supplementary information and present options for narrowing down the content (hereinafter these may be collectively referred to as "review"). It should be noted that this feedback includes not only the response by voice from the operator but also the response by text input.

At this time, for example, by grasping important phrases and keywords from the content of customer inquiries, extracting related content from existing FAQs based on this, and listening back according to the content, FAQ You can narrow it down. However, it is very burdensome for the operator to extract keywords and the like from the content of the inquiry and search the FAQ based on the extracted keywords and the like while the operator is responding to the customer. Moreover, even if the FAQ can be retrieved, it is a heavy burden to conduct appropriate feedback based on the search results, and it may be difficult to appropriately and smoothly reflect.

Therefore, the object of the present invention is to interpret the intent and meaning of inquiries from customers while eliminating as much work as possible by the operator, while guiding the operator through dialogue based on natural feedback, To provide an interactive business support system for supporting presentation of answers by supplementing and narrowing down contents.

The above and other objects and novel features of the present invention will become apparent from the description of the specification and the accompanying drawings.

A brief outline of typical inventions disclosed in the present application is as follows.

An interactive business support system according to a representative embodiment of the present invention is an interactive business support system that presents answer candidates to an operator who answers an inquiry from a customer. a text data acquisition unit that acquires text data related to an inquiry between a text data acquisition unit, a syntactic analysis unit and a semantic analysis unit that respectively perform syntactic analysis and semantic analysis of sentences related to the inquiry based on predetermined analysis rules; A dialog content recording unit that records first information based on the result of analysis by the syntactic analysis unit and the semantic analysis unit of sentences related to each inquiry recorded in the knowledge content recording unit in which the information of and have

Further, the second information based on the results of analysis by the syntax analysis unit and the semantic analysis unit for sentences related to the inquiry from the customer in the text data, and the second information recorded in the dialogue content recording unit Based on the result of matching between the first information and the second information, extracting the answer candidates recorded in the dialogue content recording unit and/or setting the contents of the reflection for specifying the answer candidates. an answer candidate extracting unit, and an information processing method that the operator uses information relating to the content of the text data relating to the inquiry from the customer, the answer candidate extracted by the answer candidate extracting unit, and the content of the set feedback. and a response support interface section for displaying in real time on the screen of the terminal.

Among the inventions disclosed in the present application, the effects obtained by representative ones are briefly described below.

That is, according to the representative embodiment of the present invention, in response to an inquiry from a customer, while eliminating as much work as possible by the operator, the operator guides the inquiry through dialogue based on a natural form of feedback. It is possible to interpret the intention/meaning of, complement/narrow down the content, and provide support for presenting an answer.

BRIEF DESCRIPTION OF THE DRAWINGS It is the figure which showed the outline|summary about the structural example of the interactive business assistance system which is Embodiment 1 of this invention. It is the figure which showed the outline|summary about the example of the screen displayed on the operator terminal in one embodiment of this invention. FIG. 4 is a flow diagram showing an overview of an example of the flow of execution system processing according to Embodiment 1 of the present invention; FIG. 2 is a diagram showing an outline of an example of interactive content according to Embodiment 1 of the present invention; FIG. FIG. 11 is a diagram showing an overview of an example of parsing processing according to Embodiment 2 of the present invention; FIG. 10 is a flow diagram showing an overview of an example of the flow of building system processing according to the second embodiment of the present invention; FIG. 11 is a flowchart showing an overview of an example of the flow of execution system processing according to Embodiment 2 of the present invention; (a) and (b) show an overview of an example in which a subject/subtitle is set for interactive content according to the second embodiment of the present invention. FIG. 10 is a diagram showing an overview of an example of a method of generating feedback according to Embodiment 2 of the present invention; FIG. 10 is a diagram showing an overview of an example of a method of generating feedback according to Embodiment 2 of the present invention;

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In principle, the same parts are denoted by the same reference numerals throughout the drawings for describing the embodiments, and repeated descriptions thereof will be omitted. On the other hand, parts that have been described with reference numerals in one drawing may be referred to with the same reference numerals, although they are not shown again in the description of other drawings.

(Overview)
The interactive business support system according to one embodiment of the present invention, for example, in a BtoC environment, enables operators to quickly and quickly perform helpdesk operations and communicator operations to immediately respond to telephone inquiries from customers. It is an information processing system that enables highly accurate processing.

FIG. 1 is a diagram showing an overview of a configuration example of an interactive business support system according to Embodiment 1 of the present invention. An interactive business support system 1 shown in FIG. 1 is a server device or the like. When there is a telephone inquiry from the customer 4 , the interactive business support system 1 analyzes voice data input via the CTI server 2 . The interactive business support system 1 outputs information based on the result of analyzing the voice data to the operator terminal 31 used by the operator 3 . The operator terminal 31 displays the information as screen information on a display unit (such as a display).

Here, before describing the details of the interactive business support system 1, an example of a screen displayed on the operator terminal 31 will be described.

FIG. 2 is a diagram showing an overview of an example of a screen displayed on the operator terminal 31 according to one embodiment of the present invention. In the example of FIG. 2, taking the case of dealing with a customer of a non-life insurance company as an example, when the operator receives a telephone inquiry from a customer and responds to it, the operator 3 is supported so that he/she can reach an appropriate answer. It shows a state in which various information for

The contents of the conversation between the customer 4 and the operator 3 are converted into text in real time by speech recognition technology and displayed in a chat format in the "contents of conversation" column in the left column of the screen. At that time, important phrases and words that are automatically extracted as keywords are highlighted.

In the "inquiry content" column in the center column of the screen, a summary sentence indicating the content of the inquiry of the customer 4 automatically recognized based on the content of the dialogue displayed in the "call content" column is displayed. In the "answer candidate" column of the column, answers extracted from existing FAQs and the like are listed as answer candidates that match the content understood as the intention of the customer's inquiry. In addition, in the "FAQ ID: ~" column in the right column of the screen, the content of the FAQ selected by the operator 3 from the list of answer candidates is displayed in preview so that the operator 3 can refer to it.

In the "additional question candidate" column in the center column of the screen, the subject ("under what circumstances can I claim insurance? ”) are automatically extracted and displayed. By specifying these items, the FAQ in the "Answer Candidates" column is further narrowed down based on the contents thereof. The operator 3 refers to this list while responding to the customer, selects an additional question for narrowing down the FAQ of answer candidates, and asks the customer 4 about the content of the additional question.

The feedback and the dialogue related to the customer 4's response to this are converted into text in real time by voice recognition technology in the same manner as above, reflected in the "content of the call" column and the "content of the inquiry" column, and the content of the answer is " It is automatically reflected in the item in the target additional question in the “Additional Question Candidates” column. For example, a toggle button (in the example in FIG. 2) for the item candidate displayed below the question content in the "additional question candidate" column (in the example in FIG. 2, for example, "Your car is covered by insurance...") , “target”, “non-target”) are automatically turned on/off according to the content of the answer (in the example of FIG. 2, the “target” button is automatically turned on).

As the contents of the "Inquiry Contents" column and the "Additional Question Candidates" column change in real time according to the contents of the reflection and the answers to this, the FAQ list in the "Answer Candidates" column is also updated moment by moment. It changes and narrows down. Then, the FAQ that finally answers the customer's inquiry is identified, and the operator can refer to this to answer the customer.

Note that when the list of FAQs in the "candidate answer" column changes from moment to moment, depending on the state of the dialogue, the operator 3 may not start from the FAQ displayed in the latest list, but from the contents of the previously displayed list. There may be cases where you want to answer (or answer) based on the FAQ that has been deleted due to the change. Therefore, for example, the FAQ currently selected by the operator 3 in the list of the "candidate answer" column and the contents of which are previewed and the FAQ specified by the operator 3 in the list can be "pinned" or "tabbed". , etc., so that the list is not deleted even if the contents of the list change.

After that, when the service to the customer 4 ends, for example, the contents of the text in the "content of the call" column are automatically summarized, and the name and other attribute information of the customer 4 grasped from the text are not shown. It may be automatically recorded as a response history in an external system such as a CRM (customer management) system or a response history management system. Further, the content of answered FAQs and the like that can be disclosed to the customer 4 may be provided to the customer 4 by e-mail or SNS (Social Networking Service).

With the above-described method, for example, manual operations such as keyword input, FAQ search, and response record creation can be eliminated and automated as much as possible. That is, the operator 3 can handle the customer 4 substantially hands-free. In the following, we will explain a method for supporting helpdesk and communicator work by extracting appropriate answer candidates while understanding the intent and meaning of inquiries through such conversations including reflection.

(Embodiment 1)
<System configuration>
Returning to FIG. 1, the details of the interactive business support system 1 will be described. The interactive business support system 1 is composed of, for example, a server device or a virtual server built on a cloud computing service as described above, and a CPU (Central Processing Unit) (not shown) controls a HDD (Hard Disk Drive) or the like. middleware such as an OS (Operating System), a DBMS (DataBase Management System), a Web server program, etc., and software running thereon are executed from the recording device. As a result, various functions, which will be described later, are implemented for supporting the operator's 3 help desk work and communicator work in response to telephone inquiries from the customer 4 .

A telephone inquiry from the customer 4 is input as voice data via the CTI server 2 . A general CTI (Computer Telephony Integration) system can be appropriately used as the CTI server 2 . It may have an IVR (Interactive Voice Response) function. An operator 3 who responds to an inquiry from a customer 4 uses an operator terminal 31, which is an information processing terminal such as a PC (Personal Computer), to refer to knowledge content 5 in which FAQs and the like are accumulated. respond while Both the CTI server 2 and the operator terminal 31 are connected so as to be able to communicate with the interactive business support system 1 via a network such as a LAN (Local Area Network) (not shown).

The interactive business support system 1 includes, for example, a syntactic analysis unit 11, a semantic analysis unit 12, an answer candidate extraction unit 13, a response support interface (I/F) 14, and a voice recognition unit 15 implemented as software. have. It also has various data such as dictionary data 16, analysis rules 17, dialogue content 18, dialogue log 19, etc. implemented by databases, files, memory tables, and the like.

The syntax analysis unit 11 has a function of performing syntax analysis such as morphological analysis on text data composed of sentences in a natural language based on contents registered in dictionary data 16 and analysis rules 17 . Also, the semantic analysis unit 12 converts the semantic attributes of the sentence of the inquiry (for example, “request”, “difficult”, “negative”, “interrogative”, etc.) into grammatical It has a function to analyze and assign For the syntactic analysis unit 11 and the semantic analysis unit 12, for example, a known morphological analysis engine, syntactic analysis engine, or the like can be used as appropriate. These pieces of information may be obtained as discourse semantics using the discourse structure analysis system described in the above-mentioned Patent Document 1.

As will be described later, the analysis by the syntax analysis unit 11 and the semantic analysis unit 12 is performed in two stages: content construction system processing (preprocessing) and execution system processing (dialogue). In the construction system processing, existing knowledge content 5 such as FAQ, which is a collection of inquiries and answers to them that have been made in the past or are expected in the future, is input and analyzed in a batch process. It outputs interactive content 18 for generating the content of feedback to the inquiry. In the execution system processing, speech data related to an inquiry from the customer 4 is converted into text by the speech recognition unit 15 and analyzed in real time by inputting the data into text, interpreting the content of the inquiry, and extracting keywords and important phrases. do.

The answer candidate extraction unit 13 interprets the intent of the inquiry based on the analysis results of the syntax analysis unit 11 and the semantic analysis unit 12 and extracts answer candidates based on the knowledge content 5 and the dialogue content 18 . A known full-text search engine may be used for extracting answer candidates. If the answer candidates can be narrowed down to one, this can be regarded as a formal answer. If the search could not be narrowed down, based on the content of the interactive content 18, additional questions (rehearsals) for narrowing down are extracted.

The response support I/F 14 has a function of displaying a screen for supporting the customer response of the operator 3 as shown in the example of FIG. 2 on the operator terminal 31 . It also has a function of recording text data relating to the content of dialogue between the customer 4 and the operator 3 as a dialogue log 19 . It is possible to obtain new entries to be added to the knowledge content 5 by separately analyzing the contents recorded in the dialogue log 19 at any time after the customer service is completed or periodically.

For example, the content recorded in the dialogue log 19 is analyzed, and a difference detection unit (not shown) compares the analysis content with the content of the knowledge content 5, and if it is new content not included in the knowledge content 5, automatically You may make it add to the knowledge content 5. FIG. Also, the analysis content and the content of the knowledge content 5 may be compared, and those having similar meanings may be grouped, and then a list may be output and periodically presented to a user such as an administrator. The user may check each content included in each group in the presented list and add a flag indicating that processing such as unification is automatically executed.

The voice recognition unit 15 has a function of converting voice data relating to the dialogue between the customer 4 and the operator 3, which is acquired from the CTI server 2 or the like, into text in real time using a known voice recognition technology. By converting not only the utterances of the inquiry about the customer 4 but also the utterances of the operator 3 into text, it is possible to grasp and analyze the entire dialogue as in the "call response" column in the screen example of FIG. In this manner, the speech recognition unit 15 acquires text data regarding inquiries between the customer 4 and the operator 3, and thus functions as a text data acquisition unit.

In particular, voice data related to the utterances of the customer 4 may be ambiguous or uninterpretable, whereas normally voice data related to the utterances of the operator 3 can be obtained clearly. Since they are educated and trained, they can clearly understand the content. Therefore, by repeating back what the operator 3 grasped about the utterance of the customer 4, such as "Isn't it ○○?" You may make it grasp the content. Whether to grasp from the customer 4's utterance or from the utterance repeated by the operator 3 may be automatically switched or may be designated by the operator 3.

<Details of processing (executive system processing)>
FIG. 3 is a flowchart showing an overview of an example of the flow of execution system processing in this embodiment. In the following example, a mobile phone company responds to an inquiry from a customer who is a subscriber. When text data relating to an inquiry from a customer is received via the speech recognition unit 15 of the interactive business support system 1, the syntax analysis unit 11 and the semantic analysis unit 12 convert the contents of the inquiry into sentences (for example, "Unsolicited mail is (S01). The comprehended content is displayed in real time, for example, in the "inquiry content" column in the screen example of FIG. 2, as described above. If the contents of the displayed inquiry are not appropriate, the operator 3 may be allowed to correct them.

When the content of the inquiry is ascertained, the text data is analyzed for its semantic content by the syntactic analysis section 11 and the semantic analysis section 12 (S02). Here, the sentence is decomposed into words by the syntactic analysis unit 11, and compound words and part-of-speech classification are performed. For example, an inquiry "I want to reject spam mail" is classified into the words "spam mail" and "reject". In order to improve the accuracy of the analysis, similar words may be standardized using the dictionary data 16 (for example, "prank mail"→"unsolicited mail").

After that, the answer candidate extraction unit 13 extracts answer candidates from the interactive content 18 (S03). Here, for example, a full-text search is performed on the interactive content 18 by a full-text search engine based on the result of analyzing the text of the inquiry in step S02. At this time, for example, the contents of the inquiry, such as specific options such as "model A", "model B", and "model C" set for the word "model" in the inquiry regarding the mobile phone It is also extracted whether or not there is an attribute value or the like set for a predetermined word as information for specifying the .

It should be noted that a list of words for which no attribute values are set may be extracted and presented to the user periodically to prompt the user to set the attribute values and the like as predetermined words. Also, when extracting words for which attribute values are not set, words with similar meanings may be grouped and presented to the user, and words with similar meanings may be automatically unified. good too.

When extracting answer candidates, for example, attribute information such as name, age, gender, address, etc. of the customer 4, purchase history, reception history, etc. are acquired from a CRM system (not shown), and these information are taken into consideration. can be For example, by referring to the purchase history, the product to be inquired about may be specified from among the products included in the purchase history. If there are a plurality of products, the operator 3 may inquire and specify them, or at that time, the candidates may be displayed in the "additional question candidate" column of the screen example of FIG. 2 to specify them. In addition, referring to past response history, if there is relevance/continuity with the content of the current inquiry, extract candidate answers after merging them automatically or according to instructions from operator 3. may

After that, the degree of certainty is calculated for each of the obtained answer candidates by a predetermined method, which will be described later, and based on this, the answer candidates are cut off and their priorities are set (S04). Then, it is determined whether or not there are a plurality of obtained answer candidates (S05). , "How to set the unsolicited mail filter of the model C...") (S06) is presented to the operator terminal 31 by the response support I/F 14. FIG. The operator 3 answers the customer 4 based on the content (and the corresponding FAQ in the knowledge content 5).

Whether or not the operator 3 actually answers the target FAQ is automatically grasped from the utterance content of the operator 3 grasped by the speech recognition unit 15, and if the operator actually answers, the knowledge The FAQ may be recorded by setting an “acceptance flag” for the FAQ of interest in the content 5 . Further, whether or not the customer 4 understands and agrees with the content can be automatically grasped from the customer 4's utterance content, and if the customer 4 understands it, it can be recorded by setting a "solution flag" or the like. good. Further, the operator 3 may be made to input information as to whether the extracted FAQ was useful or not. These pieces of information can be used, for example, as reference information when the interactive content 18 is created from the knowledge content 5 in construction-related processing, which will be described later.

On the other hand, if more than one answer candidate is obtained (S05: Yes), a predetermined method to be described later is used to ask questions (for example, (with the spam filter as a candidate) "Please tell us your model"). A candidate is generated (S07) and presented to the operator terminal 31 by the response support I/F 14. FIG. The operator 3 selects an appropriate question from the candidates for reflection according to the situation and inquires of the customer 4 about it. It is also possible to ask unique new additional questions that are not among the reflection candidates. After receiving an answer (for example, "It's model C") from the customer 4 to the feedback (S08), after that, the process from step S02 onwards is repeated as necessary with the content including the accepted answer, thereby allowing the customer to ask back. As you go along, narrow down the answer candidates.

FIG. 4 is a diagram showing an overview of an example of interactive content 18 according to this embodiment. In the example of FIG. 4, the table of the interactive content 18 has items of an ID for identifying each inquiry sentence, an inquiry sentence, reflections 1, 2, 3, . . . Each reflection item (reflection 1, 2, 3, . . . ) is provided for each content of reflection, and a keyword related to each inquiry sentence is automatically or manually set in advance in construction system processing. In addition, in the degree-of-relevance item, a pre-calculated degree of relevance of each keyword with respect to each inquiry is set. For this degree of association, for example, a known index such as TF-IDF (Term Frequency-Inverse Document Frequency) that indicates the importance of words in a sentence can be used as appropriate.

In the process of FIG. 3 described above, when extracting answer candidates by full-text search, the value of the degree of relevance item attached to the search result and whether or not each reflection item is set are also acquired. At this time, for example, a threshold value is set for the value of the degree of association, and values less than the threshold value (for example, 0.2) are cut off. In the example of FIG. 4, only the query sentences with ID=1 to 3 are extracted, and the query sentence with ID=4 (degree of relevance=0.1) is cut off. When the number of extracted query sentences is large, the query sentences may be sorted in descending order of the value of the degree of relevance, and only a predetermined number of upper-ranked queries may be extracted, and the lower-ranked query sentences may be cut off.

After that, for each of the extracted inquiry sentences, the number of reflection items for which a value is set is counted for each reflection item. In the example of FIG. 4, among the three inquiry sentences with ID=1 to 3 extracted from the interactive content 18, values are set in the item of Reflection 1 for ID=1 and 2. It is a matter. Similarly, there are 3 cases for Reflection 2 and 3 cases for Reflection 3 as well. As for Reflection 2, since all three values are "unsolicited mail" and are the same, the content of Reflection 2 is uniquely determined as "unsolicited mail".

Next, except for the uniquely determined reflection item, the relevance value for each reflection item is totaled and sorted in descending order to identify the reflection item with the highest total. In the example of FIG. 4, 0.9+0.9=1.8 for Reflection 1 and 0.9+0.9+0.3=2.1 for Reflection 3, excluding Reflection 2, which was uniquely determined. will be the highest. Then, as the processing of steps S05 and S07 in FIG. As shown in the "additional question candidate" column in the screen example of FIG. 2, a plurality of reflection candidates may be listed and displayed in descending order of the total value of the degree of relevance. The operator 3 can ask additional questions to the customer 4 based on the contents of the extracted feedback candidates.

After that, when an answer (for example, "It's model 2") is received from the customer 4, the content of the answer is automatically added to the target question in the "additional question candidate" column of the screen example in FIG. 2, as described above. reflected. The operator 3 further inquires of the customer 4 about other reflections that have not yet been asked in the "additional question candidate" column (for example, "what is the specific operation?"). By receiving answers from the customer 4 by repeatedly asking questions in this way, it is possible to acquire sufficient content related to the inquiry from the customer 4, and answer candidates from the dialogue contents 18 (and corresponding answers from the knowledge contents 5). FAQ) can be filtered dynamically.

<Details of processing (construction processing)>
In the construction system processing, interactive contents 18 to be used in an interactive conversation are generated from knowledge contents 5 such as FAQs periodically or based on instructions from the user. Here, for example, the syntactic analysis unit 11 and the semantic analysis unit 12 perform syntactic analysis and semantic analysis on each sentence (query and answer) of the knowledge content 5, extract the relationship of "object-predicate", Extract verbs and nouns as keywords to be set as reflection items. As a result, in executive processing, for example, for an inquiry consisting of verbs, it is possible to generate feedback for obtaining "what". In addition, for inquiries consisting of nouns, it is possible to generate reflections for obtaining "what to do?".

As described above, according to the interactive business support system 1 according to the first embodiment of the present invention, in response to a telephone inquiry from the customer 4, the content of the inquiry is converted into text from the voice data of the inquiry. By interpreting intentions and meanings grammatically and narrowing down candidate answers, it is possible to go beyond simple keyword matching and realize dialogues that grasp the meaning, enabling quick and highly accurate helpdesk and communicator work. is. In addition, for example, manual work such as keyword input, FAQ search, and response record creation is automated as much as possible based on voice data, and the operator 3 can respond to the customer 4 almost hands-free. can be done.

(Embodiment 2)
The interactive business support system according to the second embodiment of the present invention is slightly different from that of the first embodiment in the method of extracting answer candidates for inquiries from customers 4 and generating interactive content 18 as the basis thereof. . Since the system configuration is basically the same as the configuration shown in FIG. 1 in the first embodiment, the explanation will be omitted.

In this embodiment, as in the first embodiment, syntactic analysis and semantic analysis are performed when generating interactive content 18 and when receiving an inquiry from customer 4, and information obtained by analyzing the meaning of the inquiry text is obtained. In this embodiment, a rule for subject-analyzing sentences is added to the analysis rules 17, and "5W2H" information can be added to the results of syntactic analysis. improve. For example, assuming a service for narrowing down FAQ (knowledge content 5) as in the present embodiment, WHAT (nouns connected with wo-case and ga-case) and HOW (verbs, adjectives, and sa-transformed nouns) of 5W2H If it is possible to grasp the 1W1H of the verb conjugation), it is possible to identify the subject of the inquiry (such as "what do you want to do?") to some extent and narrow down the answer candidates with high accuracy.

FIG. 5 is a diagram showing an overview of an example of parsing processing in this embodiment. For example, the syntax analysis unit 11 parses ( The results of the morphological analysis are shown in the next row. The results of analyzing the semantic content of the inquiry by the semantic analysis unit 12 (“change (want to)”→“request”) are shown in the next stage. Furthermore, the result of analyzing the information of 5W2H (1W1H in this embodiment) based on the theme analysis rule set in the analysis rule 17 (“Plan”→WHAT, “Change”→HOW) is shown at the bottom. .

<Details of processing (construction processing)>
FIG. 6 is a flow diagram showing an overview of an example of the flow of building system processing in this embodiment. As in the first embodiment, in construction-related processing, interactive content 18 is generated from knowledge content 5 periodically or based on instructions from the user. First, the content of the knowledge content 5 is acquired (S11), and the syntactic analysis section 11 performs syntactic analysis (S12) and the semantic analysis section 12 extracts semantic attributes (S13) for each query document. After that, thematic analysis as shown in the example of FIG. 5 is performed (S14). As will be described later, there are cases where multiple subjects are included in the sentence of one inquiry. In this case, the subject (MAIN) corresponding to the essential purpose etc. and the sub-heading (SUB) corresponding to the contents of the premise etc. Identify as

Furthermore, the emotion of the customer 4 may be analyzed as necessary (S15). Emotion analysis can be performed, for example, based on matching with emotional expression patterns set in the analysis rules 17 in advance. Then, based on the above analysis results, the interactive content 18 for matching the content of the inquiry from the customer 4 with the knowledge content 5 is output (S16).

<Details of processing (executive system processing)>
FIG. 7 is a flow diagram showing an overview of an example of the flow of execution system processing in this embodiment. As in the flow chart shown in FIG. 3 of Embodiment 1, when text data relating to an inquiry from a customer is received via the speech recognition unit 15 of the interactive business support system 1, the syntax analysis unit 11 and semantic analysis The content of the inquiry is grasped as a text by the unit 12 (S21). The comprehended content is displayed in real time, for example, in the "inquiry content" column in the screen example of FIG. 2, as described above. If the contents of the displayed inquiry are not appropriate, the operator 3 may be allowed to correct them.

When the content of the inquiry is grasped, the semantic content of the text data is analyzed by the syntactic analysis unit 11 and the semantic analysis unit 12 (S22). Here, the sentence is decomposed into words by the syntactic analysis unit 11, and compound words and part-of-speech classification are performed. Synonyms may be unified using the dictionary data 16 .

After that, the answer candidate extraction unit 13 or the semantic analysis unit 12 performs subject analysis on the result of analyzing the text of the inquiry to specify 5W2H (1W1H of WHAT and HOW in this embodiment) (S24). If WHAT and HOW can be specified, that is, if the subject can be specified (S24: Yes), based on this, the answer candidate extraction unit 13 extracts answer candidates from the dialogue contents 18 (S25), It is presented to the operator terminal 31 by F14. The operator 3 replies to the customer 4 based on the contents.

On the other hand, if the subject could not be identified (S24: No), as will be described later, according to the specific situations of WHAT and HOW, the content of the feedback for identifying the unidentified subject is set ( S26), and presented to the operator terminal 31 by the response support I/F 14. FIG. The operator 3 selects an appropriate question from the candidates for reflection according to the situation and inquires of the customer 4 about it. After receiving an answer from the customer 4 to the feedback (S31), after that, the process from step S22 onwards is repeated as necessary with the content including the received answer, thereby specifying the subject of the inquiry while listening back, Narrow down the answer candidates.

<Control by Subject/Subtitle>
As described above, in the present embodiment, the theme and subtitle consisting of 1W1H of WHAT and HOW are set in advance in the interactive content 18, and the subject and subtitle are similarly specified for inquiries from the customer 4, It is possible to perform matching after interpreting the intention of the inquiry.

FIG. 8 shows an overview of an example in which a subject/subtitle is set for interactive content 18 according to the present embodiment. In FIG. 8(a), the inquiry from the customer 4 does not match the information of the subject (MAIN) for the interactive content 18 in which the subject (MAIN) and the subtitle (SUB) are set for the FAQ (knowledge content 5). FIG. 8(b) shows an example of the reflection when matching the subject (MAIN) information.

In FIG. 8A, for example, for the FAQ "I lost my mobile phone and want to stop calling", which is the content of "ID=16", as a result of theme analysis, the mere premise ""Mobile It shows that the subtitle (SUB) is set to "lost" the phone, and the main purpose is set to "stop the phone". Similarly, for the FAQ "I lost my mobile phone, so I want to cancel", which is the content of "ID = 24", "'mobile phone' is 'lost'" is set as the subtitle (SUB), and "cancellation" is set as the subject (SUB). MAIN).

Here, if the inquiry from the customer 4 is "I lost my cell phone.", parsing the syntax and unifying similar words will result in "'cell phone' is 'lost'." When this is matched with the above-described interactive content 18, the subtitle (SUB) information matches both "ID=16" and "ID=24", and both WHAT and HOW are specified, but the subject (MAIN) information does not match, ie, the subject is unknown. In this state, it is essentially unclear what the customer 4 wants to inquire about. For example, in the example of FIG. 8( a ), WHAT and HOW are specified by selectively asking each subject (MAIN) in FAQs of “ID=16” and “ID=24” with matching subtitles (SUB). ("Would you like to stop calling? Would you like to cancel?").

Further, in FIG. 8B, for example, as a result of theme analysis of the FAQ "I lost my data communication terminal", which is the content of "ID=60", "'data communication terminal' is set as the subject (MAIN). Here, if the customer's inquiry is "I lost my data communication terminal", syntactic analysis and unification of similar terms will result in "'data communication terminal' is 'lost'." . When this is matched with the interactive content 18 described above, it matches the MAIN information and specifies both WHAT and HOW. This state is a state in which the intention of the customer 4's inquiry was understood and the answer candidates were narrowed down (in the example in the figure, one was identified). 60”)) (“Sure, pause or…”).

9 and 10 are diagrams showing an overview of an example of a method for generating feedback according to the present embodiment. The reflection may be generated, for example, by the answer candidate extracting unit 13 together with the extraction of the answer candidate, or may be generated by the response support I/F 14 and displayed in the "additional question candidate" column in the screen example of FIG. You may make it

FIG. 9 shows an example of a pattern of reflection according to whether the contents of the inquiry of the customer 4 match the information of the subject (MAIN) or the subtitle (SUB) set in the interactive content 18. FIG. For example, if both the main subject (MAIN) and the sub subject (SUB) are matched, the content of the acquired main subject (MAIN) is mainly asked back, but the information obtained from the sub subject (SUB) is also confirmed as a candidate. For example, if the inquiry from the customer 4 is "I lost my mobile phone", I ask the content of the acquired subject (MAIN) again like "You lost your mobile phone, right?" repeat). In addition, it is possible to ask back the obtained subtitle (SUB) information such as "Do you wish to stop the emergency call?" or "Do you wish to terminate the call?" In addition, as for the rehearsal that is simply repeated, for example, the display in the "additional question candidate" column of the screen example of FIG. 2 may be omitted.

Also, if the subject (MAIN) matches only, the content of the acquired subject (MAIN) is asked back (recited). If there are multiple matches, confirm by selecting one of them. If only the subtitle (SUB) is matched, the contents to be obtained are confirmed based on the information of the subject (MAIN) that could not be obtained in the matched FAQ. For example, if the inquiry from the customer 4 is a "refund application form," ask back the content of the corresponding subject (MAIN) in the matching FAQ, such as "Would you like a reissue?"

Also, if neither the main subject (MAIN) nor the subheading (SUB) is matched, the fact that the information could not be acquired is reported, and confirmation is made again. For example, if the content of the inquiry from the customer 4 is ambiguous (the text result is a character string that cannot be analyzed as Japanese, etc.) (and the operator 3 himself cannot understand the content), "I'm sorry. I can't find any information that matches your question. If you don't mind, could you give me more information?"

FIG. 10 shows an example of patterns for generating sentences for reflection. In this embodiment, by combining the WHAT and HOW information of the subject (MAIN) in the FAQ that matches the inquiry, a character string for feedback is automatically generated. For example, as shown in the figure, when HOW is a noun (sa-hen noun), it is generated like "(WHAT 'no') HOW 'would you like'...". Also, when HOW is a verb, it is generated like "(WHAT 'wo') HOW 'tai'...". Also, if HOW does not exist, it is generated like "WHAT 'would you like'...". Through such reflection, it is possible to acquire the WHAT and HOW of the subject (MAIN) that have not been acquired through natural dialogue.

As described above, according to the interactive business support system 1 according to the second embodiment of the present invention, syntactic analysis and semantic analysis are performed at the time of generating the interactive content 18 and at the time of receiving an inquiry from the customer 4, Acquire information obtained by analyzing the meaning of a sentence. Furthermore, a rule for performing thematic analysis of sentences is added to the analysis rule 17 so that "5W2H (1W1H in this embodiment)" information can be added to the syntactic analysis result. As a result, it is possible to perform appropriate feedback to acquire the shortfall according to the acquisition status of WHAT and HOW related to the subject of the inquiry, and to improve the accuracy of searching for answer candidates for the interactive content 18 .

In the above-described embodiments (Embodiment 1 and Embodiment 2), the information processing system in which the operator 3 performs help desk work and communicator work for immediately responding to telephone inquiries from the customer 4 is described. rice field. Instead of this, an information processing system (for example, an information processing system having a text chat contact center) in which the operator 3 responds by text input to help desk work or communicator work that responds to inquiries by text input from the customer 4. may be applied to That is, the present invention may be applied to an information processing system in which the customer 4 side and the operator 3 side do not use voice to make inquiries and answers through text dialogue.

For example, in the above-described embodiment, the case where the speech recognition unit 15 converts the speech data of the conversation between the customer 4 and the operator 3 into text data has been described. or text data relating to the response to the inquiry input by the operator 3 may be obtained. For example, instead of including the speech recognition unit 15 described above, a text acquisition unit that acquires text input by the customer 4 and the operator 3 may be provided.

Further, whether or not the operator 3 actually answers the target FAQ is automatically grasped from the input content of the operator 3 grasped by the text acquisition unit, and if the operator actually answers, the knowledge The FAQ may be recorded by setting an “acceptance flag” for the FAQ of interest in the content 5 . Furthermore, whether or not the customer 4 understands and agrees with the content is automatically grasped from the text input content of the customer 4, and if the customer 4 understands, it is recorded by setting a "solution flag" or the like. good too. Further, the operator 3 may be made to input information as to whether the extracted FAQ was useful or not.

Although the invention made by the present inventor has been specifically described above based on the embodiments, the present invention is not limited to the above embodiments, and can be variously modified without departing from the scope of the invention. Needless to say. For example, the above embodiments have been described in detail in order to explain the present invention in an easy-to-understand manner, and are not necessarily limited to those having all the described configurations. Also, part of the configuration of one embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of one embodiment. . Moreover, it is possible to add, delete, or replace a part of the configuration of each embodiment with another configuration.

Further, each of the above configurations, functions, processing units, processing means, and the like may be realized by hardware, for example, by designing a part or all of them using an integrated circuit. Moreover, each of the above configurations, functions, etc. may be realized by software by a processor interpreting and executing a program for realizing each function. Information such as programs, tables, and files that implement each function can be stored in recording devices such as memories, hard disks, SSDs (Solid State Drives), or recording media such as IC cards, SD cards, and DVDs.

Further, in each of the above drawings, control lines and information lines are those considered to be necessary for explanation, and not all control lines and information lines for implementation are necessarily shown. In fact, it may be considered that almost all configurations are interconnected.

INDUSTRIAL APPLICABILITY The present invention can be used for an interactive business support system that supports the business of answering inquiries from customers.

1... interactive business support system, 2... CTI server, 3... operator, 4... customer, 5... knowledge content,
REFERENCE SIGNS LIST 11 syntax analysis unit 12 semantic analysis unit 13 answer candidate extraction unit 14 response support I/F 15 voice recognition unit 16 dictionary data 17 analysis rule 18 dialogue content 19 dialogue log,
31... Operator terminal

Claims (4)

An interactive business support system that presents answer candidates to an operator who answers inquiries from customers,
a text data acquisition unit that acquires text data regarding an inquiry between the customer and the operator;
a syntactic analysis unit and a semantic analysis unit that respectively perform syntactic analysis and semantic analysis on sentences related to an inquiry based on predetermined analysis rules;
first information based on the results of analysis by the syntactic analysis unit and the semantic analysis unit of sentences related to each inquiry recorded in a knowledge content recording unit storing information on inquiries and answer candidates for the inquiries; a dialogue content recording unit for recording;
second information based on the results of analysis by the syntax analysis unit and the semantic analysis unit for sentences related to the inquiry from the customer in the text data, and the first information recorded in the dialogue content recording unit; an answer candidate extracting unit for extracting answer candidates recorded in the dialogue content recording unit and setting the content of reflection for specifying the answer candidates based on the result of matching between the information of (1) and the second information; When,
The content of the text data relating to the inquiry from the customer, the answer candidate extracted by the answer candidate extracting unit, and information relating to the content of the set feedback are displayed in real time on the screen of the information processing terminal used by the operator. and a response support interface unit displayed with
The first information recorded in the dialogue content recording unit includes keywords set for each item of reflection with respect to sentences related to each inquiry, and the degree of relevance between the sentences related to the inquiry and each set keyword. including information about and
The answer candidate extracting unit performs a full-text search on each inquiry recorded in the interactive content recording unit based on the second information, and the result corresponding to the result obtained is recorded in the interactive content recording unit. Obtain the total relevance of the answer candidates with keywords set for each reflection item for the answer candidate whose relevance is above a predetermined threshold, and for the reflection item with the highest total relevance, An interactive business support system that sets the content of feedback based on keywords set as candidates. In the interactive business support system according to claim 1,
The first information recorded in the dialogue content recording unit includes WHAT information composed of nouns connected in the wo-case or ga-case for identifying the subject, analyzed for sentences related to each inquiry. , HOW information consisting of verb-like conjugations of verbs or adjectives or sa-hen nouns, and 1W1H information consisting of
The answer candidate extraction unit acquires 1W1H information by matching with the first information recorded in the dialogue content recording unit based on the second information, and acquires according to the acquisition status of 1W1H. An interactive business support system that sets the content of feedback for acquiring information related to 1W1H that has not been completed. In the interactive business support system according to claim 1,
Based on the content of the operator's response to the inquiry in the text data, the operator determines whether or not the content of the answer candidate extracted by the answer candidate extracting unit has been answered to the customer, and answers. an interactive business support system for setting information to the effect that the entry corresponding to the answer candidate stored in the knowledge content recording unit is used for the answer, if the answer candidate is stored in the knowledge content recording unit. In the interactive business support system according to claim 1,
determining whether or not the customer understands the operator's response to the content of the answer candidate extracted by the answer candidate extracting unit, based on the content of the customer's input in the text data; An interactive business support system for setting information indicating that the inquiry has been resolved to an entry corresponding to the answer candidate stored in the knowledge content recording unit when the answer candidate is understood.

Images