JP6614589B2 - Compliance check system and compliance check program - Google Patents

Description

  The present invention relates to a natural language analysis technology, and in particular, a compliance check system and a compliance check program for explaining matters that should be explained by a sales person in business negotiations with a customer and checking that matters that should not be stated are not stated. It is related to technology effective when applied to.

  For example, when a sales person in charge of a financial institution conducts a sales activity on a financial product for a customer, for compliance reasons, “Matters that must be explained (for example, risk of the financial product. "May be described as" necessary items "), and" items and expressions that should not be stated (for example, "absolutely regrets". In the following, "prohibited expressions" may be described) " It is necessary to observe the following two points: However, in most cases, sales managers (and the sales staff themselves) are not able to grasp whether this is strictly observed, and the demand for compliance is becoming stronger. In particular, there is a strong need to understand whether the above two points are being observed.

  On the other hand, for example, it is conceivable to record the contents of the business negotiations, and later check by managers to check whether the business negotiations and sales activities comply with compliance. It takes an enormous amount of time to confirm the matter, which is not realistic. In addition, for example, in the case where a business negotiation is established through a plurality of sales activities, matters that must be explained when making a proposal for a financial product in a business negotiation, for example, are necessary explanations in units of a single business negotiation. Not all. Therefore, it is necessary to confirm the contents of the business negotiations made in the past, which also increases the working time.

  As a technique for systematically grasping whether compliance is observed in business negotiations or the like, for example, Japanese Patent Laid-Open No. 2007-80096 (Patent Document 1) discloses a keyword storage for storing one or more keywords. Means, means for acquiring a voice file recording the call content, and converting the call content into data in a text searchable format based on the acquired voice file, including data in the text searchable format Means for creating a search target file, means for performing a similarity search for each keyword stored in the keyword storage means for the created search target file, and calculating the similarity for each keyword, and the calculated similarity Means for notifying a warning when there is a keyword determined to be larger than a preset value, Automatic monitoring system that allows check immediately the call contents of the sales representative and the customer are listed in the telephone business.

  In Japanese Patent No. 5449090 (Patent Document 2), a person in charge displays a document on a communication terminal used for a business talk with a customer, and records and analyzes a sound of explanation of predetermined items by the person in charge. Sales activity support that monitors whether sales activities are conducted according to the prescribed procedure by automatically determining whether the person in charge explains according to the prescribed procedure or whether an NG word is issued. The system is described.

  Japanese Patent No. 5468474 (Patent Document 3) does not relate to sales activities or business negotiations by sales representatives, but simplified discourse data in the management of operator utterances in call center operations. Processing to calculate the similarity between a block consisting of a series of utterances for each speaker and each talk script that defines an explanation for the item to be explained, and to output the most similar talk script as similarity information Talk script usage situation calculation system including is described.

JP 2007-80096 A Japanese Patent No. 5449090 Japanese Patent No. 5468474

  By using technologies such as those described in the patent literature, it is systematically determined whether or not the contents that the sales representative explained or stated to the customer during business negotiations or sales activities comply with compliance. It is possible to check to some extent.

  However, for example, in the technique of Patent Document 1, only the presence or absence of “prohibited expressions” in the content of the discourse is the target of the check, while whether “necessary items” are included is not the target of the check. In addition, the basis for checking the presence or absence of “prohibited expressions” is only the definition of the keyword, and there is a lack of flexibility.

  On the other hand, for example, in the technique of Patent Document 2, it is possible to perform both the presence / absence of “prohibited expression” and whether or not “necessary matters” are included. It can only be checked by a simple method. Even if the conversation between people talks about having the same meaning and content, the content of the utterance often fluctuates. In particular, the presence or absence of “prohibited expressions” can be checked to some extent even with the above-described relatively simple methods such as the presence or absence of NG words. Effective checking is difficult unless the contents of the negotiations that have been made are also comprehensively checked and a loose interpretation such as allowing similar expressions is taken.

  The same applies to the technique of Patent Document 3, and it is possible to systematically check whether “necessary items” are included in the content of the discourse, but it is difficult to perform an effective check as it is. In order to improve the performance, it is considered that a large amount of cost is required for the maintenance of the talk script.

  Therefore, the object of the present invention is to check whether there is a “prohibited expression” and whether “necessary items” are included based on data such as explanations to customers during business negotiations and sales activities. To provide a compliance check system and a compliance check program.

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

  Of the inventions disclosed in this application, the outline of typical ones will be briefly described as follows.

  A compliance check system according to a typical embodiment of the present invention is a compliance check system that checks whether or not each utterance made by a sales representative to a customer complies with compliance, and has the following characteristics: It is.

  That is, a text analysis unit that performs natural language analysis processing including morphological analysis on text data obtained by converting the content of each utterance of the sales representative into text using speech recognition technology, and outputs the analyzed text data, The similarity between the analyzed text data relating to each utterance and the first text data defined in advance as the content of necessary items to be explained to the customer is calculated, and the similarity exceeds a predetermined value A similarity calculation unit that determines that the necessary items of the target have been explained in the target utterance, and the analyzed text data related to each utterance of a prohibited expression that should not be described to the customer The forbidden expression of the object was stated in the utterance of the object when there was a match with the second text data predefined as content Having a keyword matching unit determines that the.

  The present invention can also be applied to a program that causes a computer to execute processing so that the computer functions as a compliance check system as described above.

  Among the inventions disclosed in the present application, effects obtained by typical ones will be briefly described as follows.

  In other words, according to a typical embodiment of the present invention, the presence or absence of “prohibited expression” and “necessary matters” are included based on data such as explanation contents to customers during business negotiations and sales activities. It is possible to check whether or not there is any.

It is the figure which showed the outline | summary about the structural example of the compliance check system which is one embodiment of this invention. It is the figure which showed the outline | summary about the example of the screen which confirms the compliance status about the negotiation etc. in one embodiment of this invention. It is the figure which showed the outline | summary about the example of the screen which confirms the compliance status about the negotiation etc. in one embodiment of this invention. It is the flowchart which showed the outline | summary about the example of the flow of the analysis process of the negotiation contents in one embodiment of this invention. It is the figure which showed the example of the specific determination of whether the "necessary matter" in one embodiment of this invention is demonstrated. It is the figure which respectively showed the outline | summary about the example of the concrete determination whether the "prohibition expression" in one embodiment of this invention is not described.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. Note that components having the same function are denoted by the same reference symbols throughout the drawings for describing the embodiment, and the repetitive description thereof will be omitted.

  The compliance check system according to one embodiment of the present invention, for example, records contents such as explanations given to customers during sales negotiations or sales activities by sales representatives such as financial institutions as voice data, and converts this into text A natural language analysis is performed based on the obtained data to check whether or not “prohibited expressions” exist and whether or not “necessary items” are included, and the result is output. As a result, the sales staff themselves and their managers can check the status of compliance in sales activities.

  In the present embodiment, the business negotiation data that has been converted into text is blocked for each of a plurality of utterances in a predetermined unit, and the fluctuation of the expression caused by the sales person explaining in the conversation with the customer is absorbed. In order to improve accuracy, synonym expansion processing is performed on the text data of the negotiations to normalize the expression. The presence or absence of “prohibited expressions” can be determined flexibly based on keyword matching, and whether or not “necessary items” are included can be determined with the text defined as “necessary items”. It is determined efficiently and effectively by calculating the degree of similarity at.

<System configuration>
FIG. 1 is a diagram showing an outline of a configuration example of a compliance check system according to an embodiment of the present invention. The compliance check system 1 includes a speech recognition server 100 and a business negotiation analysis server 200 connected to a network 300 such as the Internet, and one or more business operations including information processing terminals that can access these servers via the network 300. A terminal 411 and an administrator terminal 421 are included.

  The speech recognition server 100 is a server system that is implemented by, for example, a server device or a virtual server built on a cloud computing service, and has a function of analyzing speech data and converting it into text data. Note that the technology for converting voice data into text is not particularly limited, and known systems, software, libraries, and the like can be used as appropriate. In the present embodiment, voice data related to business negotiations and business activities is acquired from a sales terminal 411, which will be described later, via the network 300, and this is analyzed and converted into text, and the voice data and the text-converted data are also negotiated. Output to the analysis server 200.

  Similarly, the negotiation analysis server 200 is a server system implemented by, for example, a virtual server constructed on a server device or a cloud computing service, and the “prohibited expression” by keyword matching is performed on the textualized negotiation data. And whether to include “necessary items” by calculating similarity. Further, it has a function of a user interface that outputs information on negotiations together with a check result in response to a request from the administrator terminal 421 held by the administrator via the network 300.

  The negotiation analysis server 200 includes, for example, an OS (Operating System) not shown, a DBMS (DataBase Management System), a text analysis unit 210 implemented as software running on middleware such as a Web server program, a similarity calculation unit 220, a keyword Each unit includes a matching unit 230 and a user interface unit 240. Also, the negotiation voice information 201, the negotiation text information 202, the analyzed text information 203, the category assignment information 204, the synonym definition information 211, the necessary information 221, the specific expression rule 231, and the prohibited expression implemented by a database or a file Each data store has information 232 and the like.

  Here, the negotiation voice information 201 is recorded voice data acquired from the voice recognition server 100 via the network 300, and the negotiation text information 202 is a text in which the voice data is converted into text by the voice recognition server 100. Data is recorded. These data are recorded in association with information such as an ID for identifying the sales representative 410 and the customer 500 for each business negotiation so that they can be identified and extracted later.

  The text analysis unit 210 has a function of performing natural language analysis based on the text data of the negotiation recorded in the negotiation text information 202, analyzing the structure of the negotiation, and outputting the result as analyzed text information 203.

  As the natural language analysis, as will be described later, for example, morphological analysis is performed on each utterance in the text data (as described later, only those of the salesperson 410 are included, and those of the customer 500 are not included). To understand the utterance structure. The morphological analysis process can be performed using, for example, a known program or library as appropriate. Further, based on the result of the morphological analysis, for example, information related to the structure of the negotiation for grasping the meaning content of the negotiation may be acquired by determination based on a rule. For example, it is possible to grasp whether the target utterance is a question or an explanation or whether it is an unnecessary expression such as a greeting or a companion.

  In the present embodiment, in order to improve the accuracy of the check by eliminating the fluctuation of the expression in each utterance during the negotiation, as described later, based on the contents of the predefined synonym definition information 211, Normalizes expressions by expanding synonyms in text data into standard expressions.

  Whether the similarity calculation unit 220 summarizes each utterance in the analyzed text information 203 into a block made up of one or more continuous utterances according to a predetermined criterion, and whether “necessary items” are described in each block. It has a function to determine whether or not. Here, as will be described later, the degree of similarity between the text data of the utterances in each block and the text data of the explanation content of “necessary items” for each category of the negotiation item, which is defined in advance in the necessary item information 221 is calculated. . When the similarity is equal to or higher than a predetermined threshold, the “necessary items” of the target in the target block are described (the “necessary items” are included in the utterance content), and the “ The category “necessary items” is assigned and recorded in the category assignment information 204. The keyword matching unit 230 has a function of determining, by keyword matching, whether or not “prohibited expression” defined in advance in the prohibited expression information 232 is described for each utterance in the analyzed text information 203. Examples of expressions that fall under “prohibited expressions” include actual names such as names, company names, product names, telephone numbers, and addresses, such as “XXX” in “XX will always increase”. In some cases, information such as proper nouns and quantities embodied in the business negotiations (hereinafter referred to as “specific expressions”) may be included. Therefore, in the present embodiment, “prohibited expression” is defined in the prohibited expression information 232 as a parameterized text of the specific expression part.

  In this embodiment, in order to perform keyword matching with the “prohibited expression” including the above-described parameterized specific expression, specific proper nouns and quantities are determined for the text data in the analyzed text information 203. From the above, those corresponding to the specific expression are extracted in advance. Here, a rule for determining whether or not each word decomposed based on the result of the morphological analysis on the text data of the utterance is a specific expression is defined in the specific expression rule 231 as an expression pattern using a regular expression, for example. Keep it.

  The keyword matching unit 230 extracts the specific expression from the text data in the analyzed text information 203 based on the rule defined in the specific expression rule 231, and then the text of “prohibited expression” defined in the prohibited expression information 232. Keyword matching is performed using data as a keyword and taking into account specific expressions. In order to perform keyword matching here, the “prohibited expression” defined in the prohibited expression information 232 may also be defined as an expression pattern using a regular expression. If there is an expression that matches the “prohibited expression”, the category of the “prohibited expression” is assigned to the utterance and recorded in the category assignment information 204 as if the “prohibited expression” was described in the target utterance. .

  For example, in response to a request from the sales terminal 411 or the administrator terminal 421 via the network 300, the user interface unit 240 displays a list of business negotiations and their contents, and the compliance status of each business negotiation. A function to display on a web browser or the like (not shown) on the user terminal 421. Here, as will be described later, for example, the utterance information in each negotiation acquired from the negotiation text information 202 and the utterances determined to correspond to “necessary items” and “prohibited expressions” recorded in the category assignment information 204. The information is collected and displayed on the screen by displaying a list of each negotiation or displaying details of each negotiation.

  In the configuration example of FIG. 1, the voice recognition server 100 and the negotiation analysis server 200 are configured as separate server systems. However, the configuration is not limited thereto, and is configured as a single server system in which these are integrated. Also good.

  The sales representative 410 records only his / her utterance at the time of the business talk with the customer 500, so that the business staff 410 wears a directional pin microphone 412 or the like, for example, and records voice data on the sales terminal 411. Direct recording may be performed using an IC recorder or the like. The voice data is uploaded to the voice recognition server 100 at the time of reporting a daily business report to a CRM (Customer Relationship Management) system or a sales management system (not shown), for example. The utterance of the customer 500 may be recorded together, but the psychological resistance of the customer 500 is great, and the sales person 410 in the text analysis unit 210 of the speech recognition server 100 and the negotiation analysis server 200. Since it is necessary to separately process the utterance of the customer and the utterance of the customer 500, it is basically not performed.

<Screen example>
2 and 3 are diagrams showing an outline of an example of a screen that is displayed on the sales terminal 411 and the manager terminal 421 and that confirms the compliance status of business negotiations and the like. As described above, the screen is displayed by the user interface unit 240 of the business negotiation analysis server 200 on a business browser 411 or a web browser (not shown) on the administrator terminal 421 by HTML (HyperText Markup Language) or the like.

  FIG. 2 is a screen that displays a list of business negotiations, and FIG. 3 is a screen that displays detailed contents of individual business negotiations. In the example of FIG. 2, four negotiation cases are displayed in a list, and for each negotiation, information on the sales representative 410, information on the date and time of the negotiation, information on text data of utterances in the negotiation, and the like are displayed. In addition, information on a determination result as to whether or not “necessary items” are described and whether or not “prohibited expressions” are described is also displayed. For example, in the business negotiation (ID: BBB002) on the second line. The fact that “prohibited expression” is stated is highlighted together with the description “NG word: yes”.

  Further, for example, in the business negotiation (ID: DDD004) on the fourth line, “product cancellation procedure” is assigned as the “item category”, and a part of “necessary matters” to be explained in the business negotiation of the category is explained. Is displayed by the description “Description: 4/7” (4 out of 7 explanation items are explained). In this example, since the negotiation does not satisfy all the “necessary items” to be explained, the data of the negotiation is highlighted.

  Note that the “item category” information is assigned by setting by the administrator 420 or the like on the detailed screen shown in FIG. Therefore, in the example of FIG. 2, in the new case (for example, the business negotiations in the first to third lines in the figure), the “case category” is blank (or NULL), and the “case category” is still set by the administrator 420 or the like. Indicates that it has not been assigned.

  In the example of FIG. 3, detailed contents of the negotiation when the negotiation (ID: DDD004) on the fourth line in the example of FIG. 2 is selected are displayed. Text data of each utterance of the sales representative 410 is displayed. Each utterance can be confirmed by reproducing the audio data of the actual utterance content recorded in the negotiation audio information 201 by, for example, pressing a playback button consisting of a speaker mark on the right side of the column.

  Further, the upper right column indicates that the manager 420 and the like can be assigned to the negotiation by selecting or inputting a “case category” from predetermined patterns. When “Item category” is selected (in the example of FIG. 3, “Product cancellation procedure” is selected), the contents defined in advance as “necessary items” to be explained in the negotiation of the category Is extracted from the necessary information 221, and information indicating whether or not each “necessary item” is described is acquired from the category assignment information 204, and dynamically displayed in a list as illustrated.

  Here, the text data of the utterance most similar to the definition content of each “necessary item” and the similarity calculated for the utterance by the similarity calculation unit 220 (displayed as “score” in the figure) are displayed. . If the degree of similarity (“score”) is equal to or greater than a predetermined threshold, “○” is displayed in the “Description” column as if the “necessary items” have been described, and is less than the predetermined threshold. In this case, “x” is displayed as a description that is not explained.

  Also, in the lower right column, when “NG utterance”, each utterance of the sales representative 410 in the negotiation corresponds to a predefined “prohibited expression” (indicated as “NG item” in the figure). In addition, as a description of the “prohibited expression”, the text data of the utterance and the degree corresponding to the “prohibited expression” (indicated by “score” in the figure) are displayed as a list.

  The sales representatives 410 and the manager 420 can easily confirm the compliance status of the negotiations and the like through the screens shown in the examples of FIGS. 2 and 3 described above.

<Processing content>
FIG. 4 is a flowchart showing an outline of an example of the flow of analysis processing of negotiation contents in the negotiation analysis server 200. FIG. 5 is a diagram showing a specific example of whether or not “necessary items” are explained by the similarity calculation unit 220, and FIG. 6 is a “prohibited expression” by the keyword matching unit 230. It is the figure which each showed the outline | summary about the example of the concrete determination whether it is not described.

  As a premise of the processing, for example, voice data in which the sales person 410 records the contents of the negotiation with the customer 500 is uploaded to the voice recognition server 100 by the sales person 410 on the day of the negotiation. As a result, it is assumed that the negotiation voice information 201 and the negotiation text information 202 of the negotiation analysis server 200 have already been created or updated.

  In the negotiation analysis server 200, when processing is started as a night batch, first, the text analysis unit 210 acquires text data to be processed from the negotiation text information 202 (S01). 5 and 6 show examples of specific text data acquired in step S01. Here, for example, it can be obtained by directly searching the database of the negotiation text information 202 or calling an API (Application Programming Interface) created in advance.

  The text data to be processed is, for example, text data of a series of utterances for each business negotiation. Therefore, when data related to a plurality of negotiations is recorded in the negotiation text information 202, the following processing is repeated for each negotiation. As described above, since “necessary matters” may be explained through a plurality of business negotiations, information on sales negotiations 410 and customers 500 in the past is also used as a key. To get.

  Next, the text analysis unit 210 performs morphological analysis on the acquired text data (S02). Here, as described above, morphological analysis is performed using a known method, library, etc., and the part of speech of the text data is analyzed and decomposed into words (S02).

  After that, in order to eliminate fluctuations in the expression, the synonym in the text data is expanded and expressed in the standard expression based on the definition contents of the synonym definition information 211 (list of correspondence relation between the synonym and the standard expression). Is normalized (S03), and processing result information is output as analyzed text information 203. In the example of FIG. 5, for the text data acquired in step S01, in step S03, the word “domestic investment trust” is changed to the standard expression “domestic investment trust” and the word “drawn” is “reduced”. ”In the standard expression. Similarly, in the example of FIG. 6, with respect to the text data acquired in step S01, in step S03, the word “investment trust” is changed to the standard expression “investment trust”, and the word “profit” is “profitable”. It shows that it is expanded to each standard expression.

  Thereafter, it is determined whether or not “necessary items” are explained by the similarity calculation unit 220 (steps S04 to S06), and whether or not “prohibited expressions” are stated by the keyword matching unit 230 (step S07). To S09). In the example of FIG. 4, these processes are shown to be performed in parallel. However, it is not always necessary to perform these processes in parallel, and after one of the processes is completed, the other process is performed sequentially. It is good also as processing.

  In the determination process of whether or not “necessary items” are described by the similarity calculation unit 220, first, for a plurality of utterances included in the text data of the negotiation, one or more utterances continuous according to a predetermined criterion are used as a block. To summarize (S04). The example of FIG. 5 shows a state in which the initial two utterances are combined as one block in step S04. Criteria to be grouped into blocks can be, for example, a part where the explanation of the sales person 410 is cut or a part that is switched, and a time interval between consecutive utterances is longer than a predetermined length, a conjunction, a consonant, etc. Can be grouped by dividing a block with an utterance in which a predetermined word is at the beginning of the sentence.

  Next, the similarity calculation unit 220 calculates the similarity with the text data of each “necessary item” defined in the necessary item information 221 for each block acquired in step S04, and the similarity is predetermined. It is determined whether there is a “necessary item” exceeding the threshold (S05). The similarity calculation method is not particularly limited. For example, as shown in step S05 in the example of FIG. 5, the similarity appears in the text data of the target block and the text data defined as “necessary items”, respectively. It can be calculated based on the degree of matching of words of a predetermined part of speech (nouns, verbs, adjective verbs, etc.) (for example, the ratio of the number of matching words to the total number of words).

  At this time, regarding word match / mismatch, considering the entire explanation (context) of the utterance and the usual wording at that time, consider the order of multiple words and the match / mismatch of adjacencies You may do it. For example, in the example of FIG. 5, in the text data of the target block, the text data defined as “necessary items” for the adjacent relationship of “cancellation fee” → “as soon as it is handled” ”→“ Payment date ”. In this case, the word “cancellation fee” appears to match both, but it may be determined as a mismatch because the adjacent relationship is different.

  In addition, as another method generally used for determining the similarity between documents, for example, each word included in the text data of the target block and the text data defined as “necessary items”. Uses a method to calculate the importance of a word, generate a word vector with the importance as an element, and determine the degree of similarity based on the algebraic distance (or the angle formed by the word vector) between the word vectors in the vector space May be. The importance of the word here is, for example, a TF (Term Frequency) value that is an appearance frequency of a target word in a sentence, which is a commonly used index, and the target word in all sentences. A TF-IDF value that is a product of an IDF (Inverse Document Frequency) value that is the reciprocal of the number of appearing sentences can be used.

  Thereafter, when there is a “necessary item” in which the similarity calculated in step S05 exceeds a predetermined threshold, the similarity calculation unit 220 assumes that the “necessary item” has been described in the target block, and the target block For this, information specifying “necessary items” such as a category related to the “necessary items” is allocated and recorded in the category assignment information 204 (S06). If there are multiple “necessary items” that exceed a predetermined threshold, only one with the highest degree of similarity may be assigned. You may make it allocate. In the example of FIG. 5, the similarity between the text data of the target block (“Block ID: 15” in the example of FIG. 5) and the text data defined in “Category: Product Cancellation Procedure → Payment Date Explanation”. Is 71.6%, which exceeds the threshold, indicating that the category is assigned.

  On the other hand, in the determination processing of whether or not “prohibited expression” is described by the keyword matching unit 230 in steps S07 to S09, first, the text data of each utterance in the negotiation matches the content defined in the unique expression rule 231. The specific expression to be extracted is extracted (S07). In the example of FIG. 6, for the text data obtained in step S03, in step S07, the word “emerging country investment trust” is extracted as a specific expression indicating the product classification, and “<e: product classification>” is extracted. The state replaced by the parameter is shown.

  In addition, as described above, the proper expression in the present embodiment is information such as a proper noun and a quantity embodied in an actual individual negotiation such as a person name, a company name, a product name, a telephone number, and an address, that is, Although it is information that is parameterized at the time of matching, a word (such as an NG word) that directly corresponds to a “prohibited expression” may be defined by only one word.

  Next, the keyword matching unit 230 performs keyword matching with the data of each “prohibited expression” defined in the prohibited expression information 232 for the text data in which the unique expression is extracted and replaced in step S07, It is determined whether there is an utterance including “prohibited expression” (S08). In the prohibited expression information 232, for example, a rule set by an arbitrary character string or a continuous / combination of specific expressions using a regular expression or the like is defined for each category of “prohibited expressions”. By using the specific expression rule 231 and the prohibited expression information 232, a flexible “prohibited expression” can be defined.

  In the example of FIG. 6, in step S08, the second utterance out of the two utterances in the negotiation text data matches the rule defined as “category: profit guarantee” in the prohibited expression information 232. This indicates that it is determined as “prohibited expression”. At this time, the score may be calculated based on the degree of matching (the degree of matching such as words and arrangement order).

  Thereafter, the keyword matching unit 230 assigns the category information related to the “prohibited expression” to the utterance of the negotiation determined to match the “prohibited expression” rule in step S08, and records it in the category assignment information 204. (S09). If there are multiple “prohibited expressions” with matching rules for the target utterance, only one with the highest matching score may be assigned, or each “prohibited item” is described. All of these may be assigned. In the example of FIG. 6, since the text data of the target utterance (“utterance ID: 24” in the example of FIG. 6) matches the rule defined in “category: profit guarantee” with a score of 82.9%, It shows that the category is assigned.

  Returning to FIG. 4, the above-described series of processing is performed by nighttime batch processing in the negotiation analysis server 200 as described above. After that, for example, when the sales representative 410 or the manager 420 accesses the business negotiation analysis server 200 in the next day's business, a real-time process of the user interface unit 240 provides a list of business negotiations or details of individual business negotiations as requested. The contents are output and displayed (S10). Here, as described above, for example, based on the negotiation text information 202, the list information of the negotiations as shown in the example of FIG. 2 and the detailed contents information of the individual negotiations as shown in FIG. 3 are displayed on the screen. At the same time, based on the category assignment information 204, information indicating the compliance status such as the explanation status of “necessary items” in the negotiation and whether or not “prohibited expression” has been displayed is also displayed.

  As described above, according to the compliance check system 1 according to an embodiment of the present invention, the contents of the explanation etc. that the sales person 410 such as a financial institution made to the customer 500 at the time of business negotiations or sales activities. Record as audio data, perform natural language analysis based on the text data, check whether there are "prohibited expressions" and whether "necessary items" are included, and output the results Can do. As a result, the sales representative 410 itself or the manager 420 thereof can easily check the compliance status in the sales activity.

  At this time, the business negotiation data that has been made into text is blocked for each of a plurality of utterances in a predetermined unit, and the fluctuation of the expression caused by the sales representative 410 explaining in the conversation with the customer 500 is absorbed. In order to improve accuracy, synonym expansion processing is performed on the text of the business negotiations converted into text, and the expression is normalized. The presence or absence of “prohibited expressions” can be determined flexibly based on keyword matching, and whether or not “necessary items” are included can be determined with the text defined as “necessary items”. Calculate the similarity at. As a result, the accuracy of determination can be improved, and determination can be performed efficiently and effectively.

  As mentioned above, the invention made by the present inventor has been specifically described based on the embodiments. However, the present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the scope of the invention. Needless to say. For example, the above-described embodiment has been described in detail for easy understanding of the present invention, and is not necessarily limited to the one having all the configurations described. In addition, it is possible to add, delete, and replace other configurations for a part of the configuration of the above-described embodiment.

  For example, in the above embodiment, as shown in the example of FIG. 4, a plurality of utterances in the negotiation are blocked only in the determination process of whether or not “necessary items” are explained by the similarity calculation unit 220. On the other hand, the unique expression in the text data of each utterance in the negotiation is extracted and replaced only in the process of determining whether or not the “prohibited expression” is stated by the keyword matching unit 230. (S07). However, these steps are not executable only in the target determination process, and may be performed in common in both determination processes.

  The present invention can be used for a compliance check system and a compliance check program for explaining matters to be explained by a sales person in a business talk with a customer and checking that matters that should not be stated are not stated.

1 ... Compliance check system,
100 ... voice recognition server,
DESCRIPTION OF SYMBOLS 200 ... Negotiation analysis server, 201 ... Negotiation voice information, 202 ... Negotiation text information, 203 ... Analyzed text information, 204 ... Category assignment information, 210 ... Text analysis section, 211 ... Synonym definition information, 220 ... Similarity calculation section 221 ... Necessary information, 230 ... Keyword matching part, 231 ... Specific expression rule, 232 ... Prohibited expression information, 240 ... User interface part,
300 ... Network,
410 ... Sales representative, 411 ... Sales terminal, 412 ... Pin microphone, 420 ... Manager, 421 ... Manager terminal,
500 ... customer

Claims (8)

A compliance check system that checks whether the sales representatives are compliant with each utterance made to the customer,
A text analysis unit that performs natural language analysis processing including morphological analysis and outputs as analyzed text data to text data obtained by converting the content of each utterance of the sales representative into text by speech recognition technology;
The utterances in the analyzed text data related to the utterances are grouped into one or more continuous utterances according to a predetermined standard, and are defined in advance as necessary items to be explained to the customer in each block. A determination unit for determining whether or not the content of the first text data is described;
A data recording unit that records the data of each utterance made by the sales representative to the customer in association with management information that identifies the sales representative and / or the customer;
A user interface unit that displays a screen including text data obtained by converting the content of each utterance of the salesperson into text using a speech recognition technology on a terminal via a network;
Have
The text analysis unit is configured to store the utterances made to the customer by the sales person extracted based on the management information in the data of the utterances made to the customer by the sales person. Including
When the determination unit determines that the content of the first text data is explained in the block, the determination unit assigns and records the category of the necessary items to the block,
The user interface unit includes, via a network, a screen including an area for displaying text data obtained by converting the content of each utterance of the salesperson into text using speech recognition technology and an area for displaying a list of necessary items. Display it on your device,
The compliance check system, wherein the area for displaying the list of necessary items includes information in which the blocks recorded by the determination unit are associated with the necessary items. The compliance check system according to claim 1,
The text analysis unit, when the synonym defined in the synonym definition information that predefines the correspondence between the synonym and the standard expression for the synonym is included in the analyzed text data, A compliance check system that replaces the synonyms in text data with the corresponding standard expression. The compliance check system according to claim 1 or 2,
The determination unit is a compliance check system that determines a type of each utterance of the sales representative based on a predetermined utterance type determination rule set in advance based on the analyzed text data. The compliance check system according to claim 1,
The predetermined evaluation criteria is a compliance check based on the degree of matching of words of predetermined parts of speech that appear in the text data of each block and the first text data, the order of the words, or the adjacency relationship of the words, respectively. system. The compliance check system according to claim 1,
The analyzed text data related to each utterance is subject to the target text in the target utterance when there is a match with the second text data defined in advance as the content of the prohibited expression that should not be described to the customer. A compliance check system further comprising a keyword matching unit that determines that the prohibited expression is described. The compliance check system according to claim 5,
The second text data is defined as a parameter for an expression that is concrete contents at the time of utterance or a specific expression that is a word that should not be described,
The keyword matching unit is a compliance check system in which a unique expression included in the analyzed text data is extracted and parameterized, and then matched with the second text data. The compliance check system according to claim 6,
The compliance check system, wherein the second text data is defined by a sequence or combination of an arbitrary character string and / or a parameterized version of the specific expression. A compliance check program that causes a computer to execute a process so that it functions as a compliance check system that checks whether each utterance made by a sales representative to a customer complies with compliance.
Text analysis processing that outputs natural text analysis processing including morphological analysis and outputs as analyzed text data to text data obtained by converting the content of each utterance of the sales representative into text by speech recognition technology;
The utterances in the analyzed text data related to the utterances are grouped into one or more continuous utterances according to a predetermined standard, and are defined in advance as necessary items to be explained to the customer in each block. A determination process for determining whether or not the content of the first text data is described;
A data recording process for recording in association with management information identifying the sales representative and / or the customer in the data of each utterance made by the sales representative to the customer;
A user interface process for displaying a screen including text data obtained by converting the content of each utterance of the salesperson into text using a speech recognition technology on a terminal via a network;
And execute
About the past utterances made to the customer by the sales person extracted based on the management information in the data of the utterances made to the customer by the sales person in the text analysis process Including
In the determination process, when it is determined that the content of the first text data is explained in the block, the category of the necessary items is assigned to the block and recorded,
In the user interface processing, a screen having an area for displaying text data obtained by converting the content of each utterance of the sales representative into a text by a speech recognition technology and an area for displaying a list of necessary items is provided via a network. Display it on your device,
A compliance check program, wherein the area for displaying the list of necessary items includes information in which the blocks recorded in the determination process are associated with the necessary items.

Images