JP2022152061A - Information acquisition server, program, and information acquisition system - Google Patents

Abstract

To provide an information acquisition server, a program, and an information acquisition system that acquire information that can be used for operator support and business improvement from utterance content in conversation.SOLUTION: An information acquisition server 1 for acquiring information from conversation between an operator and a customer includes: a candidate word storage unit 32 that stores candidate words in association with one or more classifications; a target word acquisition unit 14 that compares a word included in a text indicating utterance content in the conversation with the candidate words in the candidate word storage unit 32 and acquires a classification target word from the text; a classification acquisition unit 15 that acquires one or more classifications of the candidate word storage unit 32 corresponding to the classification target word acquired by the target word acquisition unit 14; and an associating unit 17 that associates, when the number of the classifications acquired by the acquisition unit 15 is more than one, one classification determined by using a vicinity word located in the vicinity of the classification target word of the text with the classification target word.SELECTED DRAWING: Figure 1

Description

The present invention relates to an information acquisition server, a program, and an information acquisition system.

In a call center or the like, an operator has to respond to various customer inquiries while dealing with the customer. For example, the operator needs to sort out what the customer says and respond appropriately to the inquiry. Therefore, there is a demand for a mechanism for supporting the operator.
In such a situation, for example, recording means for recording voice data of a conversation between a customer and an operator, voice data conversion means for converting the voice data into phrase data, and the phrase data and words expressing emotions are registered in advance. Word extraction means for comparing the word list table and extracting matching words, and counting the number of the extracted words for each emotion to quantify the words expressing the emotions for each emotion, and the numerical values are stored in advance. and notification means for notifying a manager terminal that the emotion of at least one of the customer or the operator has changed remarkably when a predetermined threshold is exceeded (for example, Patent Reference 1).

JP 2016-092582 A

The device described in Patent Literature 1 assists the operator by notifying the supervisor when the operator's or the customer's emotions have significantly deteriorated.
On the other hand, as a system for supporting the operator, as described above, a system for supporting the original work of the operator, such as sorting out the content of remarks by customers and appropriately responding to inquiries, is required. In particular, when the operator is inexperienced or when the content of the inquiry is complicated, the need increases.
It is also important to analyze the content of the conversation between the operator and the customer and improve the business based on the analysis results.

SUMMARY OF THE INVENTION Accordingly, it is an object of the present invention to provide an information acquisition server, a program, and an information acquisition system for acquiring information that can be used for operator support and business improvement from the content of speech in conversation.

The present invention solves the above problems by means of the following solutions.
A first invention is an information acquisition server for acquiring information from a conversation between an operator and a customer, comprising: a candidate word storage unit storing candidate words and one or more classifications associated with each other; A target word for obtaining classification target words from the text by collating a related word storage unit storing words, words included in the text indicating the utterance content of the conversation, and the candidate words in the candidate word storage unit. acquisition means; classification acquisition means for acquiring the classification of the candidate word storage unit corresponding to the classification target word acquired by the target word acquisition means; classification determining means for determining one of the classifications by using the related words in the related word storage unit related to the classification obtained; and associating the one classification determined by the classification determining means with the classification target word. and an information acquisition server.
In a second invention based on the information acquisition server of the first invention, the classification determining means determines the one classification using a word having a dependency relation to the classification target word of the text. Acquisition server.
A third invention is the information acquisition server according to the first invention, wherein the classification determining means determines the one classification using a nearby word located near the classification target word of the text. is the server.
In a fourth invention, in the information acquisition server according to any one of the first invention to the third invention, the classification determination means determines the relationship according to the appearance frequency of words included in all the texts of the conversation. An information acquisition server that selects words to determine the one classification.
A fifth invention is the information acquisition server according to any one of the first invention to the fourth invention, wherein when there are a plurality of the classifications acquired by the classification acquisition means, the information There is an acquisition server.
In a sixth aspect based on the information acquisition server according to the first aspect, the classification determination means refers to the related word storage unit when there are a plurality of the classifications acquired by the classification acquisition means, and the A related word in the related word storage unit corresponding to one of the plurality of classifications is collated with the related word in the vicinity of the classification target word to obtain one of the related words. The information acquisition server determines the classification of the related term storage unit corresponding to the related term as the one classification.
A seventh invention is the information acquisition server according to the sixth invention, further comprising a synonym storage unit that associates and stores a representative term and a synonym of the representative term, and the classification determining means stores the one related term is not acquired, the synonym storage unit is referred to, the one related word is acquired based on the synonym that matches the neighboring word, and the related word storage corresponding to the one related word is acquired. The information acquisition server determines the classification of the part as the one classification.
An eighth invention is the information acquisition server according to any one of the first invention to the seventh invention, comprising a synonym storage unit that stores a representative term and a synonym of the representative term in association with each other; The word acquisition means refers to the synonym storage unit and extracts the text by referring to the synonym storage unit when the classification target word cannot be acquired by matching the words included in the text with the candidate words in the candidate word storage unit. is an information acquisition server that acquires the classification target term based on the synonym that matches the term included in the.
A ninth invention is the information acquisition server according to any one of the first invention to the eighth invention, further comprising clause dividing means for dividing the text into clauses, wherein the target word acquisition means is the An information acquisition server that acquires the classification target word based on the word.
In a tenth invention, the information acquisition server according to any one of the first invention to the ninth invention comprises utterance data reception means for receiving utterance data by a speaker, and the target word acquisition means accepts the utterance data. The information acquisition server acquires the classification target word from the text corresponding to the utterance data in response to receiving the utterance data.
An eleventh invention is the information acquisition server according to the tenth invention, wherein the statement data receiving means receives the text indicating the content of the statement.
According to a twelfth invention, in the information acquisition server according to the tenth invention, the utterance data receiving means receives voice data including utterance contents, and performs voice recognition processing on the voice data received by the utterance data receiving means. and a text acquisition means for acquiring the text.
A thirteenth invention is the information acquisition server according to any one of the tenth invention to the twelfth invention, comprising speaker association means for associating the text with speaker identification information for identifying the speaker. Acquisition server.
A fourteenth invention is a program for causing a computer to function as the information acquisition server according to any one of the first invention to the thirteenth invention.
A fifteenth invention is an information acquisition system comprising an information acquisition server according to any one of the first invention to the thirteenth invention, and a terminal used by the operator, wherein the information acquisition server comprises the The information acquisition system comprises output means for outputting, to the terminal, classification information including the classification and the classification target word associated by the association means.
In a sixteenth invention based on the information acquisition system according to the fifteenth invention, the information acquisition server includes an item storage unit storing in advance classification items corresponding to the requirements of the customer, and the classification associated by the association unit. and an item update output means for updating the classification items in the item storage unit corresponding to and outputting the classification items to the terminal.
In a seventeenth invention based on the information acquisition system of the sixteenth invention, the information acquisition server refers to the item storage unit based on the classification and the classification target word associated by the association means. The information acquisition system comprises requirement estimating means for estimating customer requirements, wherein the item update output means outputs the classification items corresponding to the requirements estimated by the requirement estimating means to the terminal.

According to the present invention, it is possible to provide an information acquisition server, a program, and an information acquisition system that acquire information that can be used for operator support and business improvement from the content of statements in a conversation.

1 is an overall configuration diagram of an information acquisition system and a functional block diagram of an information acquisition server according to this embodiment; FIG. It is a figure which shows the example of the memory|storage part of the information acquisition server which concerns on this embodiment. It is a figure which shows the example of the memory|storage part of the information acquisition server which concerns on this embodiment. 7 is a flowchart showing information acquisition processing of the information acquisition server according to the embodiment; 6 is a flowchart showing text classification processing of the information acquisition server according to the embodiment; This is a continuation of FIG. FIG. 6 is continued. It is a figure which shows the example of a process of the information acquisition server which concerns on this embodiment. It is a figure which shows the example of a process of the information acquisition server which concerns on this embodiment. It is a figure which shows the example of the memory|storage part of the information acquisition server which concerns on this embodiment. It is a figure which shows the example of a display screen in the operator terminal which concerns on this embodiment.

EMBODIMENT OF THE INVENTION Hereafter, the form for implementing this invention is demonstrated, referring a figure. This is just an example, and the technical scope of the present invention is not limited to this.
(embodiment)
FIG. 1 is an overall configuration diagram of an information acquisition system 100 and a functional block diagram of an information acquisition server 1 according to this embodiment.
2 and 3 are diagrams showing examples of the storage unit 30 of the information acquisition server 1 according to this embodiment.

<Information acquisition system 100>
In the information acquisition system 100 shown in FIG. 1, for example, the information acquisition server 1 receives utterance data in a conversation between a customer and an operator at a call center from a customer terminal 4 and an operator terminal 5. FIG. The information acquisition system 100 is a system in which the information acquisition server 1 classifies the words included in the utterance data, thereby acquiring information that can be used for operator support and business improvement from the utterance content of the conversation.

The information acquisition system 100 includes an information acquisition server 1, a customer terminal 4 of each customer, and an operator terminal 5 (terminal) of each operator. The information acquisition server 1, the customer terminal 4, and the operator terminal 5 are connected via a communication network N so as to be able to communicate with each other. Although FIG. 1 shows one customer terminal 4 and one operator terminal 5 , there may be a plurality of customer terminals 4 and operator terminals 5 , each of which is connected to the information acquisition server 1 .
The communication network N, which will be described later, is a data communication network in the case of exchanging text between the customer terminal 4 and the operator terminal 5. FIG. In addition, the communication network N includes, for example, a voice communication network in addition to the data communication network in the case where voice communication is performed between the customer terminal 4 and the operator terminal 5 .

In the following embodiment, the information acquisition system 100 will be described as an example in which an operator answers a customer's inquiry about insurance. However, the business that can use the information acquisition system 100 is not limited to this.

<Information acquisition server 1>
The information acquisition server 1 is a device for acquiring information by classifying words included in utterance data indicating utterances between a customer and an operator. The information acquisition server 1 is owned by, for example, a company that operates a call center or a company entrusted by the call center. The information acquisition server 1 may be composed of one server, may be composed of a plurality of servers, or may be a cloud, for example.

The information acquisition server 1 includes a control section 10 , a storage section 30 and a communication interface section 39 .
The control unit 10 is a central processing unit (CPU) that controls the entire information acquisition server 1 . The control unit 10 reads and executes an operating system (OS) and application programs stored in the storage unit 30 as appropriate, thereby cooperating with the above-described hardware and executing various functions.
The control unit 10 includes a statement data reception unit 11 (statement data reception means, text acquisition means), a speaker association unit 12 (speaker association unit), a segmentation unit 13 (segment segmentation means), and a target word acquisition unit. 14 (target word acquisition means), classification acquisition section 15 (classification acquisition means), classification determination section 16 (classification determination means), association section 17 (association means), item update section 18 (requirement estimation means, item and an information output unit 19 (output means, item update output means).

The utterance data reception unit 11 receives utterance data from the speaker from the customer terminal 4 and the operator terminal 5 .
The utterance data reception unit 11 may receive text utterance data, which is a message, from the customer terminal 4 and the operator terminal 5 using, for example, a web chat or the like in which a customer and an operator transmit messages in real time. .
Further, the utterance data receiving unit 11 may receive utterance data, which is voice data, from the customer terminal 4 and the operator terminal 5, for example, using the function of the telephone. In the case of voice data, for example, a call acquisition device (not shown) installed between an operator's telephone (not shown) and an operator's microphone (not shown) is used to capture the speech of both the customer and the operator. It is possible to acquire the audio data of the content. Then, the utterance data receiving unit 11 performs voice recognition processing on the received voice data, and acquires text indicating the content of the utterance.

The speaker association unit 12 associates the text acquired by the message data reception unit 11 with identification information (speaker identification information) such as an ID (IDentification) for identifying a speaker. When the utterance data receiving unit 11 receives text directly from the customer terminal 4 and the operator terminal 5, the speaker association unit 12 may associate the text with an ID corresponding to the sender, for example. Further, when the utterance data receiving unit 11 receives voice data, the speaker association unit 12 may assign a different ID based on the input direction of the voice data to the call acquisition device and associate it with the text. . Furthermore, for example, when a plurality of customers are speaking at the customer terminal 4, different IDs are assigned based on the characteristics of the speech data for each speech, the directivity of the speech based on the position of the speaker, the volume, etc. May be given and associated with text.
Then, the speaker association unit 12 associates the text with the ID and stores it in the conversation storage unit 37 .

The phrase division unit 13 divides the text received by the utterance data reception unit 11 into phrases, and acquires words from the phrases. Words obtained from phrases are, for example, words.
The target word acquisition unit 14 collates the words included in the phrases of the text acquired by the phrase division unit 13 with the candidate words in the candidate word storage unit 32 to acquire classification target words from the text. Here, if there is a candidate word in the candidate word storage unit 32 that matches a word included in the text, the target word acquisition unit 14 acquires the candidate word as a classification target word. Further, when there is no candidate word in the candidate word storage unit 32 that matches the word included in the text, the target word acquisition unit 14 compares the word included in the text with synonyms in the synonym storage unit 34. to see if there are any matching synonyms. Then, when there is a matching synonym, the target word acquiring unit 14 acquires a representative word corresponding to the synonym as a classification target word.
Communication between the customer and the operator is colloquial, especially in voice conversations. Therefore, even words with the same meaning are assumed to use various expressions and words. Therefore, the target word acquisition unit 14 uses the synonym storage unit 34 to deal with fluctuations in expression.
The classification acquisition unit 15 acquires the classification of the candidate word storage unit 32 corresponding to the classification target word acquired by the target word acquisition unit 14 .

The classification determination unit 16 selects the classification target words acquired by the target word acquisition unit 14 and the related words in the related word storage unit 33 related to the classification of the candidate word storage unit 32 corresponding to the classification target words acquired by the classification acquisition unit 15. and determine the classification (one classification). More specifically, the classification determining unit 16 may determine the classification using words having a dependency relationship with respect to the classification target word acquired by the target word acquiring unit 14, for example. Further, the classification determination unit 16 may determine the classification using, for example, neighboring words located near the classification target word acquired by the target word acquisition unit 14 . Here, the term "neighboring word" means, for example, a word to be acquired within a predetermined definition such as how many words the word is within the classification target word. Further, the classification determination unit 16 may select related words in the related word storage unit 33 according to the appearance frequency of words included in the text, and determine one classification. Furthermore, the classification determination unit 16 may perform processing by the classification determination unit 16 when, for example, there are a plurality of classifications acquired by the classification acquisition unit 15 .

For example, when the classification acquisition unit 15 acquires a plurality of classifications, the classification determining unit 16 selects a neighboring word, which is a word located near the classification target word of the text, and one of the plurality of classifications. One related word is acquired by matching with the related words in the related word storage unit 33, and the classification of the related word storage unit 33 corresponding to the one related word is determined to be one classification.
Further, when there is a related word in the related word storage unit 33 that matches the neighboring word and corresponds to one of a plurality of classifications, the classification determining unit 16 classifies the related word into one related word. The classification of the related word storage unit 33 corresponding to one related word acquired as a word is determined as one classification. On the other hand, if there is no related word in the related word storage unit 33 that matches the nearby word and there is no related word corresponding to any of the plurality of classifications, the classification determining unit 16 34 synonyms to see if there is a matching synonym. Then, when there is a matching synonym, if the representative term corresponding to the synonym is a related term corresponding to one of a plurality of classifications, the classification determining unit 16 treats the representative term as one related term. The classification of the related word storage unit 33 corresponding to one related word is determined to be one classification.

The association unit 17 associates the classification acquired by the classification determination unit 16 with the classification target word. Then, the association unit 17 stores the classification information of the associated classification and the classification target word in the association storage unit 38 .

The item update unit 18 refers to the item storage unit 35 and guesses the customer's requirements based on the classification and the classification target word associated by the association unit 17 . Here, the item updating unit 18 may perform the estimation based on, for example, the sufficiency rate of the category items in the item storage unit 35 according to the categories associated by the association unit 17 . Then, the item update unit 18 updates the value of the classification item in the item storage unit 35 corresponding to the inferred requirement to the classification target word corresponding to the classification associated by the association unit 17 .
The information output unit 19 outputs, to the operator terminal 5 , classification information including the classification and the classification target word associated by the association unit 17 . The information output unit 19 also outputs the values of the classification items in the item storage unit 35 updated by the item update unit 18 to the operator terminal 5 .

The storage unit 30 is a storage area such as a hard disk or a semiconductor memory device for storing programs, data, etc. necessary for the control unit 10 to execute various processes.
The storage unit 30 includes a program storage unit 31, a candidate word storage unit 32, a related word storage unit 33, a synonym storage unit 34, an item storage unit 35, a conversation storage unit 37, and an association storage unit 38. Prepare.
The program storage unit 31 is a storage area that stores various programs. The program storage unit 31 stores programs for performing various functions executed by the control unit 10 of the information acquisition server 1 .

The candidate word storage unit 32 is a storage area that associates and stores candidate words and classifications.
As shown in FIG. 2A, the candidate word storage unit 32 stores one or more candidate words in association with one or more classifications. Here, a candidate word is a word that is a candidate for determining the classification, which simply indicates the content of the statement. Also, the classification is a category indicating the content of the candidate word. For example, there are two classifications corresponding to candidate words such as "hospitalization insurance" and "surgery insurance": "subject to billing" and "subject to cancellation". Further, for example, a classification corresponding to candidate words such as “insurance claim” and “document confirmation” is one of “procedure details”. Candidate words and classifications are stored in advance in the candidate word storage unit 32 in association with each other.

The related word storage unit 33 is a storage area for storing classifications and related words in association with each other.
As shown in FIG. 2B, the related word storage unit 33 stores one or more related words in association with one classification. Here, a related word is a word related to classification. Classifications and related words are stored in advance in the related word storage unit 33 in association with each other.

The synonym storage unit 34 is a storage area that stores representative words and synonyms in association with each other.
As shown in FIG. 3A, the synonym storage unit 34 stores one synonym in association with one representative word. Here, a representative word indicates a word used as a candidate word or a related word. Note that the synonym storage unit 34 may correspond to one representative word with a plurality of synonyms. The synonym storage unit 34 stores in advance representative words and synonyms in association with each other. Further, the synonym storage unit 34 may be able to be updated by a predetermined registration work when a new synonym is generated.

The item storage unit 35 is a storage area that stores requirements and classification items in association with each other.
As shown in FIG. 3B, the item storage unit 35 associates requirements with a plurality of classification items and stores them. Here, a requirement is a category indicating the content of a customer's inquiry. Also, the classification item is an item that needs to be obtained from the customer, and corresponds to the classification. In the item storage unit 35, requirements and classification items are stored in advance in association with each other.

The conversation storage unit 37 is a storage area that associates and stores the ID of the speaker and the text of the content of the speech. In response to the control unit 10 accepting the statement data and specifying the speaker, for example, the ID of the speaker and the text of the content of the statement are registered in the conversation storage unit 37 as needed. In addition, the conversation storage unit 37 associates and stores the ID of the speaker and the text of the content of the speech for each unit from the start to the end of the conversation between the customer and the operator.

The association storage unit 38 is a storage area that associates and stores classifications and extraction values. Here, the extracted value is a classification target word obtained from the text corresponding to the classification, or a representative word of the word obtained from the text corresponding to the classification. The association storage unit 38 associates and stores the classification and the extracted value, with the start to end of the conversation between the customer and the operator as one unit.

Note that the above-described storage units of the storage unit 30 are merely examples, and may be stored by different holding methods. For example, the candidate word storage section 32 and the related word storage section 33 may be integrated, or the contents of the synonym storage section 34 may be included. Also, the storage unit 30 is not limited to this, and may have other storage areas such as a storage area for storing information about the operator.
The communication interface unit 39 is an interface for performing data communication and voice communication with other devices via the communication network N. FIG.

<Customer terminal 4>
A customer terminal 4 shown in FIG. 1 is a terminal used by a customer. A customer uses the customer terminal 4 to make an inquiry to the operator.
The customer terminal 4 is, for example, a mobile terminal having computer functions, such as a smart phone. The customer terminal 4 may be a personal computer (PC), a tablet terminal, or the like, or may be a telephone.
The customer terminal 4 includes, although not shown, a control section, a storage section, an input section, a display section, a communication interface section, and the like. The customer terminal 4 may be provided with a touch panel display in which an input section and a display section are integrated.

<Operator terminal 5>
An operator terminal 5 shown in FIG. 1 is a terminal used by an operator. The operator uses the operator terminal 5 to answer inquiries from customers.
The operator terminal 5 is, for example, a PC. The operator terminal 5 may also have a telephone and a microphone.
The operator terminal 5 includes, although not shown, a control section, a storage section, an input section, a display section, a communication interface section, and the like.

Here, the computer means an information processing apparatus having a control section, a storage device, etc., and the information acquisition server 1, the customer terminal 4, and the operator terminal 5 are information processing apparatuses each having a control section, a storage section, and the like. Yes, included in the concept of a computer.

<Processing of Information Acquisition Server 1>
Next, processing of the information acquisition server 1 will be described.
FIG. 4 is a flowchart showing information acquisition processing of the information acquisition server 1 according to this embodiment.
5 to 7 are flowcharts showing text classification processing of the information acquisition server 1 according to this embodiment.

For example, when the control unit 10 of the information acquisition server 1 detects that the customer terminal 4 has connected to the information acquisition server 1, the information acquisition process shown in FIG. 4 is started.
In step S (hereinafter, "step S" is simply referred to as "S") 11 in FIG. 4, the control unit 10 (utterance data receiving unit 11) receives the utterance data. Here, the utterance data may be text or voice data. When the utterance data is voice data, the control unit 10 (utterance data receiving unit 11) performs voice recognition processing on the voice data and acquires text indicating the contents of the utterance. In addition, the speech data includes both that of the customer and that of the operator. Further, utterance data refers to data from the start to the end of utterances by any of the speakers, and may be one sentence or multiple sentences.

In S12, the control unit 10 (speaker association unit 12) assigns an ID for identifying the speaker of the received utterance data to the text.
In S<b>13 , the control unit 10 (speaker association unit 12 ) associates the ID with the text and stores them in the conversation storage unit 37 .
In S14, the control unit 10 performs text classification processing. The text classification process is a process of obtaining the classification of words included in the text. The information acquisition server 1 can organize the statement content by acquiring the word classification.

Here, text classification processing will be described with reference to FIGS. 5 to 7. FIG.
In S21 of FIG. 5, the control unit 10 (clause dividing unit 13) divides the text into clauses. Then, the control unit 10 (clause delimiter 13) acquires the words included in the clause. Here, when there are multiple words included in the phrase, the control unit 10 may acquire multiple words.
In S<b>22 , the control unit 10 (target word acquiring unit 14 ) compares the acquired word with candidate words in the candidate word storage unit 32 . Here, matching with the candidate words may be performed for all of the acquired words at once, or may be performed sequentially from the first word of the text.
In S23, the control unit 10 (target word acquiring unit 14) determines whether or not the candidate word has been matched. If the candidate word is matched, that is, if there is a candidate word in the candidate word storage unit 32 that matches the acquired word (S23: YES), the control unit 10 shifts the process to S24. On the other hand, if the candidate word cannot be matched (S23: NO), the control unit 10 shifts the process to S25.

In S24, the control unit 10 (target word acquisition unit 14) acquires the word that has been matched with the candidate word as a classification target word. After that, the control unit 10 shifts the processing to S31 in FIG.
In S<b>25 , the control unit 10 (target word acquiring unit 14 ) collates the word acquired in the process of S<b>21 with synonyms in the synonym storage unit 34 .
In S26, the control unit 10 (target word acquisition unit 14) determines whether or not synonyms have been matched. If the synonym is matched, that is, if there is a synonym in the synonym storage unit 34 that matches the acquired word (S26: YES), the control unit 10 shifts the process to S27. On the other hand, if synonyms cannot be matched (S26: NO), the control unit 10 moves the process to S15 of FIG. 4 on the condition that all the words included in the clause have been processed. If there is an unprocessed word included in the phrase, the control unit 10 shifts the process to S22 and sequentially processes the unprocessed word.
In S27, the control unit 10 (target word acquisition unit 14) acquires representative words corresponding to the synonyms in the synonym storage unit 34 as classification target words. After that, the control unit 10 shifts the processing to S31 in FIG.

In S31 of FIG. 6, the control unit 10 (classification acquisition unit 15) refers to the candidate word storage unit 32 and extracts the classification of the candidate word corresponding to the acquired classification target word. determine whether there is If there is one classification (S31: YES), the control unit 10 shifts the process to S32. On the other hand, if there is more than one classification (S31: NO), the control unit 10 shifts the process to S33. A case where there is more than one classification means a case where there are a plurality of classifications.
In S<b>32 , the control unit 10 (classification determining unit 16 , associating unit 17 ) associates the extracted classification with the classification target word and stores them in the association storage unit 38 . After that, the control unit 10 shifts the process to S15 in FIG. 4 on the condition that all the words included in the clause have been processed.

In S<b>33 , the control unit 10 (classification determining unit 16 ) collates neighboring words, which are words near the classification target word, with related words in the related word storage unit 33 .
In S<b>34 , the control unit 10 (classification determination unit 16 ) determines whether or not matching with related terms is possible. If the related word is matched, that is, if there is a related word in the related word storage unit 33 that matches the nearby word (S34: YES), the control unit 10 shifts the process to S35. On the other hand, if the related term could not be matched (S34: NO), the control unit 10 shifts the process to S37 in FIG.
In S35, the control unit 10 (classification determination unit 16) acquires the related words that have been matched as one related word.
In S<b>36 , the control unit 10 (the classification determination unit 16 and the association unit 17 ) associates the classification corresponding to the acquired one related word with the classification target word, and stores them in the association storage unit 38 . After that, the control unit 10 shifts the process to S15 in FIG. 4 on the condition that all the words included in the clause have been processed.

In S<b>37 of FIG. 7 , the control unit 10 (classification determination unit 16 ) compares neighboring words, which are words near the classification target word, with synonyms in the synonym storage unit 34 .
In S38, the control unit 10 (classification determination unit 16) determines whether or not synonyms have been matched. If synonyms can be matched, that is, if there is a synonym in the synonym storage unit 34 that matches the near term (S38: YES), the control unit 10 shifts the process to S39. On the other hand, if synonyms cannot be matched (S38: NO), the control unit 10 moves the process to S15 of FIG. 4 on the condition that all the words included in the clause have been processed.
In S39, the control unit 10 (classification determination unit 16) acquires the representative term of the synonym storage unit 34 corresponding to the synonyms that have been collated as one related term.
In S<b>40 , the control unit 10 (the classification determination unit 16 and the association unit 17 ) associates the classification corresponding to the acquired related word with the classification target word, and stores them in the association storage unit 38 . After that, the control unit 10 shifts the process to S15 in FIG. 4 on the condition that all the words included in the clause have been processed.

In S15 of FIG. 4, the control unit 10 determines whether or not the classification has been acquired. If the classification has been acquired (S15: YES), the control unit 10 shifts the process to S16. On the other hand, if the classification could not be acquired (S15: NO), the control unit 10 shifts the process to S18.
In S16, the control unit 10 (item update unit 18) performs item update processing. Specifically, the item updating unit 18 refers to the item storage unit 35 based on the classifications and classification target words stored in the association storage unit 38 to infer the customer's requirements. Then, the item update unit 18 updates the value of the classification item in the item storage unit 35 corresponding to the inferred requirement to the classification target word corresponding to the classification associated by the association unit 17 .

In S<b>17 , the control unit 10 (information output unit 19 ) performs output processing to the operator terminal 5 . Specifically, the information output unit 19 outputs the classification information including the classification and the classification target words stored in the association storage unit 38 to the operator terminal 5 . The information output unit 19 also outputs the values of the classification items in the item storage unit 35 to the operator terminal 5 .
In S18, the control unit 10 determines whether or not the speech has ended. For example, when the control unit 10 of the information acquisition server 1 detects that the customer terminal 4 has disconnected from the information acquisition server 1, the control unit 10 determines that the speech has ended. If the speech has ended (S18: YES), the control unit 10 ends this process. On the other hand, if the speech has not ended (S18: NO), the control unit 10 shifts the process to S11.

<Specific example>
Next, a specific example using the above-described processing will be described.
8 and 9 are diagrams showing an example of processing of the information acquisition server 1 according to this embodiment.
FIG. 10 is a diagram showing an example of the storage unit of the information acquisition server 1 according to this embodiment.
FIG. 11 is a diagram showing an example of a display screen on the operator terminal 5 according to this embodiment.

<Specific example 1>
FIG. 8A shows an example of the text 41 of the utterance data received from the customer terminal 4. FIG.
The control unit 10 divides the text 41 into clauses and acquires the words 42 (S21 in FIG. 5). Then, the control unit 10 collates the word 42 with the candidate words in the candidate word storage unit 32 (S22 in FIG. 5).
FIG. 8B shows the collation result. As shown in FIG. 8B, since the candidate word 51 matches the word 42, the control unit 10 acquires the word 42 as a classification target word (S24 in FIG. 5).
Next, as shown in FIG. 8(B), there are two classifications of candidate words corresponding to the classification target word: "subject to billing" and "subject to cancellation" (NO in S31 of FIG. 6). As shown in FIG. 8(C), 10 acquires the neighboring words 43 in the vicinity of the word 42, and compares the neighboring words 43 with the related words in the related word storage unit 33 (S33 in FIG. 6). ).

FIG. 8D shows the collation result. As shown in FIG. 8D, there is no related word that matches the neighboring word 43 (NO in S34 of FIG. 6). Therefore, the control unit 10 next compares the neighboring term 43 with synonyms in the synonym storage unit 34 (S37 in FIG. 7).
FIG. 8(E) shows the collation result. As shown in FIG. 8(E), since the synonym 52 matches the neighboring term 43, the control unit 10 acquires the representative term 53 corresponding to the synonym 52 as one related term (S39 in FIG. 7). ).
Then, as shown in FIG. 8(F), the control unit 10 acquires the classification 55 corresponding to one related word 54 in the related word storage unit 33, and, as shown in FIG. It is associated with the word 42 and stored in the association storage unit 38 (S40 in FIG. 7).
In this way, the information acquisition server 1 can sort out the contents of the statement of the text 41 by the classification and extraction value stored in the association storage unit 38 .

<Specific example 2>
FIG. 9A shows an example of text 61 of utterance data received after text 41 from customer terminal 4 .
The control unit 10 divides the text 61 into clauses and acquires words 62 (S21 in FIG. 5). Note that the control unit 10 acquired the word "claimant" before the word 62 and performed the text classification process (S14 in FIG. 4), but could not acquire the classification. Then, the control section 10 matches the word 62 with the candidate words in the candidate word storage section 32 (S22 in FIG. 5). Although not shown, since there is no candidate word that matches the word 62 (NO in S23 of FIG. 5), the control unit 10 next compares the word 62 with synonyms in the synonym storage unit 34 (see FIG. 5). 5 S25).

FIG. 9B shows the collation result. As shown in FIG. 9B, the synonym 71 matches the word 62 (YES in S26 of FIG. 5), so the control unit 10 acquires the representative word corresponding to the synonym 71 as the classification target word 72. (S27 in FIG. 5).
Next, since there is one classification of candidate words corresponding to the classification target word 72 as shown in FIG. 9C (YES in S31 of FIG. The classification 73 corresponding to the candidate word is acquired, and as shown in FIG. 9D, the classification 73 and the classification target word 72 are associated and stored in the association storage unit 38 (S32 in FIG. 6).

<Specific example 3>
Concrete example 3 shown in FIG. 10 is a conversation stored in the storage unit 30 as a result of a conversation between a customer and an operator, including the concrete example 1 described based on FIG. 8 and the concrete example 2 described based on FIG. Examples of a storage unit 37 and an association storage unit 38 are shown.

<Specific example 4>
Concrete Example 4 shown in FIG. 11 shows an example of the operator screen 80 output to the operator terminal 5 during the conversation between the customer and the operator in Concrete Example 3 shown in FIG.
Operator screen 80 includes a conversation area 81 and a support area 91 .
The conversation area 81 is an area related to conversation with the customer. Conversation area 81 includes a statement output area 82 and a statement input area 83 .
The statement output area 82 outputs the content of statements between the operator and the customer in chronological order and in a manner in which the speaker can be identified.
The comment input area 83 is for the operator to input comments to the customer, and includes an input field 84 and a send button 85 .

The support area 91 is an area for outputting information in which the contents of remarks are organized in order to support the operator. The support area 91 includes a classification information area 92 and a candidate area 93 .
The classification information area 92 outputs the classification information stored in the association storage section 38 .
Candidate area 93 outputs a category item for which no value has been entered among the category items in item storage unit 35 . Although the classification items are output together with the priority in FIG. 11, the priority may be output in the order in which the classification items in the item storage unit 35 are arranged, for example. Also, the priority may not be output.
By referring to the support area 91 during the conversation with the customer, the operator can grasp the content of the customer's inquiry while looking at the output of the information obtained from the customer after sorting it out. can. In addition, it is possible to confirm the items to be asked to the customer, and to have an efficient conversation with the customer.
Note that the operator screen 80 shown in FIG. 11 is an example. Other ways of presentation may be used, and, for example, a screen of only the support area 91 may be used.

Thus, according to the information acquisition server 1 of this embodiment, the following effects are obtained.
(1) Each storage unit of the storage unit 30 is referenced from the utterance content of the conversation to acquire a classification for organizing the utterance content. For example, the word to be classified is acquired from the word of the text and the candidate word storage unit 32 that stores the candidate word and one or more classifications in association with each other. Next, the classification corresponding to the classification target word is determined using the related words in the related word storage unit 33 that stores the related words related to the classification. Then, the determined classification and the classification target word are associated. Therefore, it is possible to acquire information by increasing the accuracy of extracting the classification for the words of the text. In addition, it is possible to use related words to check whether the utterances in the conversation are related to the classification target word and the classification.

(2) when a word in the text corresponds to a plurality of classifications, obtains one related word by comparing the related word storage unit 33 storing related words related to the classification with neighboring words; Since the classification is determined from one related word that has been acquired, it is possible to obtain the classification related to the word in the text from the neighboring words, thereby improving the accuracy of the classification.

(3) Since the synonym storage unit 34 is used to acquire one related word or classification target word, fluctuations in various expressions used in conversation can be dealt with.
(4) Since the text is divided into clauses and the words included in the clauses are acquired as words to be classified, the words for information extraction can be obtained easily and appropriately.
(5) When the utterance data is voice data, it is converted into text by voice recognition processing. Therefore, even if the utterance data is voice data, information corresponding to the classification can be extracted. In addition, since it is possible to deal with the case where the utterance data is text, for example, it is possible to deal with inquiries from customers through Web chat.

(6) Since the classification information including the classification and the association with the words to be classified is output to the operator terminal 5, the operator can grasp the customer's inquiry at a glance based on the classification information in which the contents of the inquiry from the customer are organized. It is convenient to do.
(7) The classification items corresponding to the customer's requirements are stored in advance in the item storage section 35, and the classification items in the item storage section 35 are updated in accordance with the association between the classification and the classification target word. Since the information is output to the operator terminal 5, it is possible to assist the operator so that he/she does not miss the contents to be confirmed by the customer.
In addition, since the customer's requirements are estimated from the update status of the classification items in the item storage unit 35 and the classification items of the item storage unit 35 corresponding to the estimated requirements are output to the operator terminal 5, the operator's requirements including the customer requirements are output. can support

Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments. Moreover, the effects described in the embodiments are merely enumerations of the most suitable effects produced by the present invention, and the effects of the present invention are not limited to those described in the embodiments. The above-described embodiments and modifications described later can be used in combination as appropriate, but detailed description thereof will be omitted.

(deformed form)
(1) In the present embodiment, when there are multiple classifications in the candidate word storage unit corresponding to a word in the text (classification target word), the related words in the related word storage unit are used as an example. Not limited. Even if there is only one classification in the candidate word storage unit corresponding to the classification target word, related words in the related word storage unit may be used. In doing so, the accuracy of text classification can be improved.

(2) In the present embodiment, an example has been described in which neighboring words of text words (classification target words) are compared with related words, but the present invention is not limited to this. A related word may be collated with a word having a dependency relationship with respect to the classification target word. By doing so, the accuracy of text classification can be further improved by using words that are highly related to the classification target word.
Also, the classification may be determined by selecting related words according to the appearance frequency of words contained in all texts of the conversation. In this way, the classification can be determined using related words that are common throughout the conversation, so that a consistent classification can be obtained throughout the conversation.

(3) In the present embodiment, an example has been described in which the operator is assisted by processing in real time upon receipt of utterance data and outputting classification information to the operator terminal 5, but this is the case. is not limited to The utterance data stored in the conversation storage unit may be collectively processed. By doing so, the service provider can analyze the content of inquiries not only for one conversation but also for inquiries from a plurality of customers. The analysis results can be used, for example, to improve and enrich FAQs (Frequently Asked Questions) and to improve the content for sales promotion.

(4) In the present embodiment, text words (classification target words) are associated with classification, but the present invention is not limited to this. The text itself may be associated with the classification.
(5) In the present embodiment, an inquiry from a customer regarding insurance has been described as an example, but the present invention is not limited to this. It can also be used for other call center operations (inquiries, invitations, etc.).

1 information acquisition server 10 control unit 11 utterance data reception unit 12 speaker association unit 13 clause division unit 14 target word acquisition unit 15 classification acquisition unit 16 classification determination unit 17 association unit 18 item update unit 19 information output unit 30 storage unit 31 program Storage unit 32 Candidate word storage unit 33 Related word storage unit 34 Synonym storage unit 35 Item storage unit 37 Conversation storage unit 38 Association storage unit 39 Communication interface unit 80 Operator screen 81 Conversation area 91 Support area 92 Classification information area 93 Candidate area 100 Information acquisition system

Claims (17)

An information acquisition server that acquires information from a conversation between an operator and a customer,
a candidate word storage unit that stores candidate words in association with one or more classifications;
a related word storage unit that stores related words related to the classification;
target word acquisition means for acquiring a classification target word from the text by comparing words included in the text indicating the utterance content of the conversation with the candidate words in the candidate word storage unit;
Classification acquisition means for acquiring the classification of the candidate word storage unit corresponding to the classification target word acquired by the target word acquisition means;
classification determining means for determining one of the classifications using the classification target words of the text and the related words in the related word storage section related to the classification acquired by the classification acquisition means;
an association means for associating the one classification determined by the classification determination means with the classification target word;
Information acquisition server. In the information acquisition server according to claim 1,
The information acquisition server, wherein the classification determining means determines the one classification using a word having a dependency relationship with respect to the classification target word of the text. In the information acquisition server according to claim 1,
The information acquisition server, wherein the classification determining means determines the one classification using a nearby word located near the classification target word of the text. In the information acquisition server according to any one of claims 1 to 3,
The information acquisition server, wherein the classification determination means selects the related words according to the appearance frequency of words included in all the texts of the conversation and determines the one classification. In the information acquisition server according to any one of claims 1 to 4,
An information acquisition server for performing processing by the classification determination means when there are a plurality of the classifications acquired by the classification acquisition means. In the information acquisition server according to claim 1,
When there are a plurality of classifications acquired by the classification acquisition means, the classification determination means refers to the related term storage unit to determine the proximity words in the vicinity of the classification target word of the text and the plurality of classifications. The related words in the related word storage unit corresponding to any of the classifications are collated to acquire one related word, and the classification in the related word storage unit corresponding to the one related word is stored in the An information acquisition server that determines one classification. In the information acquisition server according to claim 6,
A synonym storage unit that stores representative words and synonyms of the representative words in association with each other,
When the one related term cannot be acquired, the classification determination means refers to the synonym storage unit, acquires the one related term based on the synonym that matches the neighboring term, and An information acquisition server that determines the classification of the related word storage section corresponding to the one related word to the one classification. In the information acquisition server according to any one of claims 1 to 7,
A synonym storage unit that stores representative words and synonyms of the representative words in association with each other,
The target word acquisition means refers to the synonym storage unit when the classification target word cannot be acquired by matching the words included in the text with the candidate words in the candidate word storage unit, An information acquisition server that acquires the classification target term based on the synonym that matches the term included in the text. In the information acquisition server according to any one of claims 1 to 8,
Clause delimiting means for segmenting the text into clauses;
The information acquisition server, wherein the target word acquisition means acquires the classification target word based on the word included in the clause. In the information acquisition server according to any one of claims 1 to 9,
utterance data receiving means for receiving utterance data by a speaker;
The information acquisition server, wherein the target word acquiring means acquires the classification target word from the text corresponding to the utterance data in response to the utterance data receiving means receiving the utterance data. In the information acquisition server according to claim 10,
The utterance data receiving means is an information acquisition server that receives the text indicating the content of the utterance. In the information acquisition server according to claim 10,
The utterance data receiving means receives voice data including utterance content,
An information acquisition server comprising text acquisition means for performing speech recognition processing on the voice data received by the utterance data reception means and acquiring the text. In the information acquisition server according to any one of claims 10 to 12,
An information acquisition server, comprising a speaker association unit that associates the text with speaker identification information that identifies the speaker. A program for causing a computer to function as the information acquisition server according to any one of claims 1 to 13. an information acquisition server according to any one of claims 1 to 13;
a terminal used by the operator;
An information acquisition system comprising
The information acquisition system, wherein the information acquisition server comprises output means for outputting, to the terminal, classification information including the classification and the classification target word associated by the association means. In the information acquisition system according to claim 15,
The information acquisition server is
an item storage unit storing in advance classification items corresponding to the customer's requirements;
item update output means for updating the classification item of the item storage unit corresponding to the classification associated by the association unit and outputting the classification item to the terminal;
An information acquisition system comprising: In the information acquisition system according to claim 16,
The information acquisition server includes requirement inference means for inferring requirements of the customer by referring to the item storage unit based on the classification and the classification target word associated by the association means,
The information acquisition system, wherein the item update output means outputs the classification item corresponding to the requirement estimated by the requirement estimation means to the terminal.

Images