KR102431754B1 - Apparatus for supporting consultation based on artificial intelligence - Google Patents

Abstract

An artificial intelligence-based counseling support apparatus according to an embodiment of the present invention includes an input unit into which consultation contents are input, an STT conversion unit that converts voice information input through the input unit into text, derives keywords from the converted text, and provides a counselor's terminal The central processing unit that runs the counseling app with It includes an artificial intelligence reasoning unit that recommends

Description

Apparatus for supporting consultation based on artificial intelligence}

The present invention relates to an artificial intelligence-based counseling support apparatus, and more particularly, to an apparatus in which a counselor can receive consultation content support using artificial intelligence.

Even today, with the development of computing devices such as artificial intelligence, it is clear that direct counseling by counselors is an indispensable customer service channel for consumers in relation to economic activities in society as a whole, such as banks and insurance companies.

However, as the number of knowledge increases exponentially day by day, consulting based on the intellectual ability of one or two counselors with a vast amount of information is at risk of receiving inaccurate knowledge from the point of view of the service provider as well. A situation is coming in which the burden of work becomes that much greater.

Recently, there has been a device that helps in counseling by using a real-time search tool or extracting usable content from a database that stores similar counseling history in counseling.

It is necessary to introduce an artificial intelligence-based device that can help counselors to conduct counseling through the existing past history, and at the same time provide customized information for sales tasks and reasoning on customer questions.

Republic of Korea Patent Registration No. 10-1901920 (A device and method for providing reverse transcription service between voice and text for artificial intelligence voice recognition deep learning) is a prior art that introduced artificial intelligence to a counseling device.

However, the above prior art uses artificial intelligence to change spoken language into text and uses artificial intelligence technology to connect appropriate counselors according to the customer's consultation content. do not perform

In addition, the content of counseling supported by counselors must maintain a high level of trust, and the need to introduce artificial intelligence into counseling support devices in order to quickly respond to changing information is growing day by day.

It is an object of the present invention to provide an artificial intelligence-based counseling support device for supporting counseling contents of a counselor to a customer's request for counseling.

An artificial intelligence-based counseling support apparatus according to an embodiment of the present invention includes an input unit into which consultation contents are input, an STT conversion unit that converts voice information input through the input unit into text, derives keywords from the converted text, and provides a counselor's terminal The central processing unit that runs the counseling app with It includes an artificial intelligence reasoning unit that recommends

Here, the central processing unit is a login module that controls access of a customer or a counselor through input log-in information, a customer information storage module that stores the customer's personal information and consultation history, and a consultation that executes a program for consultation with the terminal of the counselor It includes a keyword extraction module for extracting keywords by analyzing the text input through the app providing module and the STT conversion unit.

Here, the information retrieval management unit includes a content storage module for storing counseling content information in a hierarchical structure divided into upper hierarchical information and lower hierarchical information, and a search engine module for performing information retrieval through information input by the counselor.

Here, the artificial intelligence inference unit uses a query hierarchy analysis module that determines a hierarchy of customer query contents using keywords provided from the central processing unit, and keywords provided from the central processing unit and hierarchical information provided from the query hierarchy analysis module. and a learning data providing module that provides learning data for customer query contents.

In addition, the artificial intelligence inference unit can be recommended using a dialog model storage module that stores input text and conversation contents in a hierarchical structure divided into upper hierarchical information and lower hierarchical information, and customer's personal information, counseling history, and input text. It may further include a product recommendation module for inferring a product.

In addition, the learning data providing module performs an information search with the search engine module of the information search management unit when the learning data search in the dialogue model storage module fails.

Here, the learning data may be the consultation content information or the conversation content provided for each layer in the conversation model storage module.

The artificial intelligence-based counseling support apparatus according to an embodiment of the present invention can provide accurate information to a counselor in real time.

In addition, it is possible to build an improved system without changing the device through interworking with the existing knowledge management system.

In addition, real-time sales tasks such as new product recommendations are possible at the same time as consulting with customers, thereby contributing to the creation of performance of counselors.

1 is a structural diagram of a counseling system according to an embodiment of the present invention.
2 is a functional block diagram of a counseling support apparatus according to an embodiment of the present invention.
3 is a detailed block diagram of a central processing unit according to an embodiment of the present invention.
4 is a detailed block diagram of an information retrieval management unit according to an embodiment of the present invention.
5 is a detailed block diagram of an artificial intelligence inference unit according to an embodiment of the present invention.
6 is a customer consultation screen of a counselor according to an embodiment of the present invention.

The features and effects of the present invention described above will become more apparent through the following detailed description in relation to the accompanying drawings, and accordingly, those of ordinary skill in the art to which the present invention pertains can easily implement the technical idea of the present invention. will be able

Since the present invention can have various changes and can have various embodiments, specific embodiments are illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the present invention to a specific embodiment, it should be understood to include all modifications, equivalents and substitutes included in the spirit and scope of the present invention.

In describing each figure, like reference numerals are used for like elements.

Terms such as first, second, etc. may be used to describe various elements, but the elements should not be limited by the terms. The above terms are used only for the purpose of distinguishing one component from another.

For example, without departing from the scope of the present invention, a first component may be referred to as a second component, and similarly, a second component may also be referred to as a first component. and/or includes a combination of a plurality of related listed items or any of a plurality of related listed items.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related art, and should not be interpreted in an ideal or excessively formal meaning unless explicitly defined in the present application. shouldn't

The suffix module, block, and part for the components used in the following description are given or mixed in consideration of only the ease of writing the specification, and do not have distinct meanings or roles by themselves.

Hereinafter, a preferred embodiment of the present invention will be described with reference to the accompanying drawings so that those of ordinary skill in the art can easily implement it. In the following description of embodiments of the present invention, if it is determined that a detailed description of a related known function or a known configuration may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted.

The present invention is a device applied to a counseling system that provides easy-to-access counseling services to customers in banks, insurance companies, securities companies, service centers, and the like. A consultation support apparatus according to an embodiment of the present invention relates to a device for storing consultation content with a customer, and performing learning based on the accumulated consultation content to recommend a consultation item suitable for the customer's needs at the right time and place.

1 is a structural diagram of a counseling system according to an embodiment of the present invention.

As shown in FIG. 1 , in general, a customer requests a consultation using various channels such as telephone consultation, chatting consultation, and video consultation. The customer is connected to the counselor through the call infrastructure 20 , and the counseling support device 10 according to an embodiment of the present invention recommends counseling content suitable for the customer's counseling content to the counselor.

Here, the counselor may include both manned/unmanned counseling, and the counseling support apparatus 10 according to the embodiment of the present invention may be connected to the counselor through the call infrastructure 20 .

The call infrastructure 20 is a PBX (Private Branch eXchange) that connects the customer's terminal and the agent's terminal, and IVR (Interactive Voice Response) that performs call processing control for selectively connecting the customer's terminal to the agent's terminal connected to the local network , STT (Speak To Text), etc. may be included.

2 is a functional block diagram of a counseling support apparatus according to an embodiment of the present invention.

As shown in FIG. 2 , the counseling support apparatus 10 according to an embodiment of the present invention, which recommends counseling contents to a counselor in a counseling system in real time, through an input unit 100 into which the counseling contents are input, and the input unit 100 STT conversion unit 200 that converts input voice information into text, a central processing unit 300 that derives a keyword from the converted text, and runs a counseling app with a counselor's terminal, provides search information matching the keyword, It includes an information retrieval management unit 400 for storing consultation contents and an artificial intelligence reasoning unit 500 for analyzing the converted text and recommending an answer to the counselor's query or the customer's query through machine learning.

The input unit 100 is a device for receiving voice information input through a terminal of a customer and voice information input through a terminal of a counselor. Also, the input unit 100 may be a device for receiving text input through a terminal of a customer or a counselor during consultation through chatting. That is, all information input to the counseling support device 10 is transmitted to the counseling support device 10 through the input unit 100 . The input unit 100 may include not only hardware devices such as a microphone and keyboard, but also wired/wireless telecommunication lines and communication means for transmitting input signals to the central processing unit of the counseling support apparatus.

The STT conversion unit 200 includes a conversational separation module 210 for separating the call contents of each of the customer and the counselor, a recorder module 220 for recording the speaker's call contents, and a documentation module 230 .

In this specification, a module may mean a functional and structural combination of hardware for carrying out the technical idea of the present invention and software for driving the hardware. For example, the module may mean a logical unit of a predetermined code and a hardware resource for executing the predetermined code, and does not necessarily mean physically connected code or one type of hardware.

The dialogue separation module 210 functions to separate the dialogue voice between the counselor and the customer.

As the consultation work of call centers becomes routine, when incoming and outgoing calls between the counselor and the customer are connected, the contents of the consultation are introduced to the customer and then the recording is carried out. Recorded files are used as evidence in case of customer complaints or as sample selection and supporting data for counselor evaluation. As the services that use the audio recording file in the consultation work are diversified, there is a limit to the use of the recording file of a single channel that does not distinguish between the counselor and the customer.

Various methods can be used for speaker separation, segmenting the voice data based on the silence section, extracting the MFCC (Mel-Frequency Cepstral Coefficient) feature vector, and separating the speakers through K-mean clustering based on this. technology may be used, but is not necessarily limited thereto.

The recorder module 220 functions to store the conversation voice of the customer in the dialog separation module 210 . Of course, it is also possible to store the counselor's conversational voice, and if necessary, it can be used to record the actual conversation without separating the conversationalist.

The STT conversion unit 200 according to an embodiment of the present invention includes a documenting module 230 for automatically converting text by recognizing a dialogue voice.

The documentation module 230 is a functional block that implements the actual speech recognition technology. In general, after extracting the sound signal, the noise is removed, and the features of the speech signal are extracted and compared with the speech model database (DB) for speech recognition. can do.

3 is a detailed block diagram of a central processing unit according to an embodiment of the present invention.

As shown in FIG. 3 , the central processing unit 300 according to an embodiment of the present invention is a device for operating the consultation support device 10 , and includes a login module 310 , a customer information storage module 320 , and a consultation app providing module. 330 and a keyword extraction module 340 .

The login module 310 performs customer matching in customer information through the ID and password input by the user. When there is no customer information matching the input ID and password, a display window for inducing membership registration is output to the customer's terminal as in the general case.

The customer information storage module 320 stores personal information, such as name, resident number, phone number, address, subscription product, ID, password, and the like, which the customer entered at the time of membership registration. When the customer information storage module 320 is connected to the customer's terminal and an ID and password are input, it matches with the stored personal information to provide data for customer identification.

In addition, the customer information storage module 320 stores all available information such as a history of prior consultation, a consultation method, a history searched in the site, and a history of visiting a website.

The counseling app providing module 330 may be a customer counseling program provided to the counselor's terminal. In particular, a dedicated screen for customer consultation as shown in FIG. 6, which will be described below, is provided, and various menus can be freely used.

The consultation app providing module 330 may provide different programs according to various consultation methods such as phone consultation, chat consultation, video consultation, etc., and further provides a program capable of managing the counselor's activities and managing consultation contents. .

The keyword extraction module 340 extracts a keyword from the customer's consultation contents converted into text through the STT conversion unit 200 . Keywords are extracted mainly from terms used in related industries, but key words determined by frequency of occurrence and similarity with related industries are extracted.

The keyword extraction module 340 may extract at least one keyword from the conversation content of the customer, and is classified into keywords included in each layer according to the layer information stored in the information search management unit 400 to be described below. may be extracted.

4 is a detailed block diagram of an information retrieval management unit according to an embodiment of the present invention.

As shown in Fig. 4, the information retrieval management unit 400 uses the information input by the counselor and the content storage module 410 to store the consultation content information in a hierarchical structure divided into upper hierarchical information and lower hierarchical information. and a search engine module 420 for performing information retrieval.

When the content storage module 410 stores, for example, information related to savings products, individual savings products (such as savings savings for first-timers in their 20s, happy life savings savings, college life savings savings, etc.) constitute upper layer information, and the lower layer As information, for each installment savings product, the individual items corresponding to each savings account are subordinated to 1. Product characteristics, 2. Subscription target, 3. Saving method and savings amount, 4. Interest rate and interest rate, 5. Product type, 6. Additional service. hierarchical information is formed.

In addition, another lower layer information may be added to the lower layer information, and the content storage module 410 may connect and store a plurality of layer information with each other.

The search engine module 420 performs a search with text input in the search window of the counseling app provided to the counselor. Through the intranet of the company, the internal database can be utilized or a search can be performed based on the public web.

The artificial intelligence inference unit 500 according to the embodiment of the present invention utilizes various contents stored in the information search management unit 400 in the central processing unit 300 to support consultation, and is not determined in advance through self-learning, or to ask questions. Instead of providing formal content with a fixed answer to the question, it allows the counselor to check inferable or predicted question content or additional answer content related to the consultation content.

5 is a detailed block diagram of an artificial intelligence inference unit according to an embodiment of the present invention.

As shown in FIG. 5 , the artificial intelligence inference unit 500 according to an embodiment of the present invention uses a keyword provided from the central processing unit 300 to determine a hierarchy of customer query content 510 . , a learning data providing module 520 that provides learning data for customer query contents using keywords provided from the central processing unit 300 and hierarchical information provided from the query hierarchy analysis module 510, input text and A dialogue model storage module 530 that stores the conversation contents in a hierarchical structure divided into upper hierarchical information and lower hierarchical information, and a product recommendation module that infers recommendable products using the customer's personal information, consultation history, and input text ( 540).

The query hierarchy analysis module 510 determines to which hierarchy the customer's query content belongs by using the keywords provided from the keyword extraction module 340 of the central processing unit 300 . The learning data providing module 520 provides learning data by using the extracted keyword and hierarchical information. Here, the learning data is the conversation content provided by each layer in the consultation content information or the dialog model storage module. In particular, the counseling content information included in the learning data may be information belonging to the same layer that the customer may be interested in. That is, the consultation content information may be information of the same layer with a high frequency of additional questions or a high possibility of additional queries through learning in the dialogue model stored in the dialogue model storage module 530 among information belonging to the same layer. In addition, the conversation contents included in the learning data may be conversation contents including contents belonging to the same hierarchical information among the dialogue models from which the same keyword is extracted, and by providing this to the counselor, the conversation contents with high probability to be developed in the future can be predicted in advance. You can inform your counselor.

The learning data providing module 520 provides information through the search engine module 420 of the information retrieval management unit 400 when the search for the counseling content information or the learning data including the conversation content fails in the dialog model storage module 530 . Search can be performed to provide structured content.

Looking in detail through the preceding example, the upper layer information, which is information about the type of savings savings, is stored in the content storage module 410 as savings savings for first-timers in their 20s, happy life savings, college life savings, and the like.

In addition, for each savings product, which is detailed sub-tier information by type of savings account, 1. product characteristics, 2. subscription target, 3. accumulation method and savings amount, 4. interest rate and interest rate, 5. product type, 6. additional services, etc. information is stored in the content storage module 410 .

If the keyword extracted as the customer query content is "Happy Life Savings Savings", the query layer analysis module 510 determines that it corresponds to upper layer information, and the learning data providing module 520 determines "Happy Life Savings Savings" among the information in the same layer. It is possible to provide the counselor with other savings products of interest to the subscribers of "Happy Life Savings Savings" through machine learning or savings products that are highly relevant to the counselor as learning data.

In addition, conversation contents belonging to the same hierarchical information in the past and matching keywords can be provided to the counselor as learning data.

On the other hand, if the keywords extracted from the customer query are "Happy Life Savings Savings" and "Saving Amount", among the information related to lower-level information, "interest rate and interest rate", "receivable amount", etc. information can be provided to the counselor as learning data.

When the keyword extracted through the keyword extraction module 340 does not exist or cannot be extracted, the learning data providing module 520 induces the counselor to extract a keyword that meets the needs of the customer through a query.

The dialog model storage module 530 stores the input text and dialog content in a hierarchical structure like the content storage module 410 described above. The keywords of "Happy Life Savings Savings" are extracted and the conversation between the counselor and the customer is the conversation content corresponding to the upper layer information, tag data is added and stored, and "Happy Life Savings Savings" and "Saving Amount" are derived as keywords. The conversation contents are stored in the dialogue model storage module 530 after tag data is added to the dialogue contents as lower-level information.

The product recommendation module 540 infers recommendable products using the customer's personal information, consultation history, and input text, and provides them to the counselor. Considering the gender, age, occupation, etc. of the customer who is consulting with the counselor, and extracting the products additionally subscribed by customers with similar personalities or histories through the extracted keywords through machine learning, and inferred recommendable products information to the counselor.

6 is a customer consultation screen of a counselor according to an embodiment of the present invention.

As shown in FIG. 6 , the customer's consultation contents converted into text through the STT conversion unit 200 are output through the chatting window. When the customer (Hong Gil-dong) asks the counselor a query, “I am curious about the interest rate of savings products for student finances,” the keyword extraction module 340 of the central processing unit 300 uses “student savings savings” and “interest rate” as keywords. extract In the consultation content information item in the lower right corner, pre-stored standardized information related to the interest rate of the student loan savings savings stored in the content storage module 410 is output so that the counselor can immediately check and answer.

In addition, in the recommended learning data items at the bottom of the chat window, other information of the lower layer information, which is the layer to which “Student fund preparation savings (upper layer information)” and “interest rate (lower layer information)” belong, “monthly payment information”, “expiration Additional items that customers may be curious about, such as “information on received amount” and “information on early termination fee,” are recommended to counselors as learning data through query hierarchy analysis, etc. You can confirm the contents of the query you can ask. Such learning data may be recommendations for other items in the same layer, but may also be recommendable product information as described above.

That is, through the counseling support device 10 according to the embodiment of the present invention, the counselor immediately responds to the customer's inquiry, apart from providing a formally determined answer to the customer's inquiry. This is possible, and it is possible to create the performance of the consultant through sales marketing for other recommended products.

According to the software implementation, not only the procedures and functions described in this specification but also the design and parameter optimization for each component may be implemented as a separate software module. The software code may be implemented as a software application written in a suitable programming language. The software code may be stored in a memory and executed by a controller or a processor.

10 Consultation support device 20 Call infrastructure
100 input unit 200 STT conversion unit
210 dialog separation module 220 recorder module
230 Documentation module 300 Central processing unit
310 Login module 320 Customer information storage module
330 Consulting app provision module 340 Keyword extraction module
400 Information retrieval management unit 410 Content storage module
420 search engine module 500 artificial intelligence reasoning unit
510 Query layer analysis module 520 Learning data provision module
530 dialog model storage module 540 product recommendation module

Claims (3)

an input unit into which consultation contents are input;
STT conversion unit for converting the voice information input through the input unit into text;
a central processing unit for deriving keywords from the converted text and executing the counseling app on the counselor's terminal;
an information search management unit that provides search information matching keywords and stores consultation contents; and
It includes an artificial intelligence reasoning unit that analyzes the converted text and recommends answers to the agent's query or customer's query through machine learning,
The information retrieval management unit
a content storage module for storing consultation content information in a hierarchical structure divided into upper hierarchical information and lower hierarchical information; and
Includes a search engine module that performs information search through the information input by the counselor,
The artificial intelligence inference unit
In addition to supporting counseling, we do not provide formal content that is not predetermined through self-learning or that has a fixed answer to a question, but provides inferable or predicted question content or additional answer content related to the consultation content,
a query hierarchy analysis module for determining a hierarchy of customer query contents using keywords provided from the central processing unit;
a learning data providing module for providing learning data in the same layer for customer query contents using keywords provided from the central processing unit and hierarchical information provided from the query hierarchical analysis module;
a dialog model storage module for storing input text and conversation contents in a hierarchical structure by adding tag data to distinguish them into upper hierarchical information and lower hierarchical information; and
A product recommendation module for inferring recommendable products using the customer's personal information, consultation history, and input text;
The learning data providing module is
Counseling support apparatus, characterized in that the information search is performed from the search engine module of the information search management unit when the learning data search in the dialogue model storage module fails. According to claim 1,
The central processing unit
a login module for controlling access of customers or consultants through input login information;
Customer information storage module for storing the customer's personal information and consultation history;
Consulting app providing module for executing a program for consultation with the counselor's terminal; and
and a keyword extraction module for extracting keywords by analyzing the text input through the STT conversion unit. According to claim 1,
The learning data is
Counseling support apparatus, characterized in that the counseling content information or the conversation content provided by layer in the dialog model storage module.

Images