KR102241532B1 - Intelligent callbot server and unmanned counsel systeim using thereof - Google Patents

Abstract

The intelligent call bot server and the unmanned counseling method using the same according to an embodiment of the present invention is a STT server capable of converting a customer's voice into a received text through an STT server and improving the speech recognition rate based on the analysis result of the received text. The type may be determined, and voice recognition may be performed by connecting an optimal STT server among the STT servers according to the determined type of the STT server. At this time, the optimal STT server may be determined using any one of the customer's speech type, speech topic, and speech intention.

Description

Intelligent callbot server and unmanned counseling method using it {INTELLIGENT CALLBOT SERVER AND UNMANNED COUNSEL SYSTEIM USING THEREOF}

The present invention relates to an intelligent call bot server and an unmanned counseling method using the same, and more particularly, to an intelligent call bot server that enables unmanned counseling for a customer accessing a call center service, and an unmanned counseling method using the same.

In general, many companies, such as manufacturers and service providers, operate customer centers such as call centers to provide customer service for products they sell or services they provide.

Conventional call centers use a voice guidance system to connect customers and counselors to receive counseling information from customers, and counselors directly enter the received counseling information into a computer, and select a person in charge to handle the service content. The subsequent procedure was carried out.

However, the conventional call center system has a problem in that work efficiency decreases and labor expenses increase due to the need for a counselor to perform many tasks. In order to overcome these limitations, a call bot system has been introduced and used that can conduct counseling without a counselor.

The call bot system conducts conversations according to pre-built scenarios or learning contents, converts the customer's voice to text using the STT (Speech To Text) engine, determines the intention based on the converted text, and selects the answer content. After the output comment is sent to the customer by voice using the TTS (Text To Speech) engine.

It is very important for such a call bot system to quickly and accurately analyze the customer's voice data to extract the correct text. However, according to the characteristics of the STT engine, the accuracy of the speech recognition result is different, and the speech data is transmitted in a fixed packet size, so that the accuracy of speech recognition is deteriorated.

In the case of Korean Patent Application Laid-Open No. 10-2016-0100412, it discloses starting the voice recognition processing immediately after the start of the call, but does not disclose a configuration capable of improving the accuracy and speed of voice recognition.

Korean Patent Publication No. 10-2016-0100412 (2016.08.23)

The present invention is to solve the above problems, and in detail, an object of the present invention is to provide an intelligent call bot server capable of improving the speed and accuracy of voice recognition, and an unattended counseling method using the same.

The problems of the present invention are not limited to the problems mentioned above, and other problems that are not mentioned will be clearly understood by those skilled in the art from the following description.

The intelligent call-bot server according to an embodiment of the present invention for solving the above problems is interlocked with an exchange of a call center system, a TTS server, and a plurality of STT servers to provide an unattended consultation service to a customer, but the customer's consultation through the exchange When a request is received, a call is connected to the exchange, the customer's voice input to the exchange is converted into a received text through the STT server, the converted received text is analyzed to generate an outgoing text, and the outgoing text Is configured to convert the outgoing voice into an outgoing voice through the TTS server, and transmit the outgoing voice to the customer through the exchange, and determine the type of STT server capable of improving the voice recognition rate based on the analysis result of the received text. , It may include an STT server determination unit for performing voice recognition by connecting an optimal STT server among the STT servers according to the determined type of the STT server.

The STT server determination unit may analyze the received text to determine any one of a speech type, speech topic, and speech intention to determine a type of STT server capable of improving a speech recognition rate.

The STT server determination unit has previously built an STT type database regarding type information of the optimal STT server for each speech topic, and may select and connect an STT server that is optimal for the speech topic based on the STT type database.

The STT server determination unit may select and connect to an AI-based STT server when the utterance form is a natural language form or when it is determined that the utterance intention includes a plurality of intentions.

The STT server determination unit may select and connect a specialized STT server when the utterance type is a specialized utterance form including a terminology or when it is determined that the utterance intention includes a single intention.

In addition, the intelligent call-bot server according to an embodiment of the present invention analyzes the customer's voice data packet to calculate a silent section or a quiet section in which a volume value is less than a preset reference value, and uses a buffer to calculate the silent section or It may further include a packet processing unit for varying a packet unit before and after the attenuating period and transmitting it to the STT server.

In addition, the intelligent call bot server according to an embodiment of the present invention further includes a call processing unit for managing a conversation history and voice recording between the customer and the intelligent call bot server, the call processing unit to the counselor terminal connected through a network. A conversation history may be provided in real time, and link information for confirming the voice recording may be provided.

The call processing unit may be configured to provide link information for confirming the conversation history and the voice recording to the customer's terminal connected through a network.

The call processing unit may provide link information for confirming the voice recording when the customer's identity verification is completed.

In addition, the intelligent call bot server according to an embodiment of the present invention includes a scenario management unit that provides a scenario management screen for real-time confirmation and change of a conversation scenario between the customer and the intelligent call bot server to a counselor terminal connected through a network, The scenario management screen may be configured to form a flow by connecting a plurality of nodes having different functions in a drag-and-drop manner.

In addition, an unattended counseling method using an intelligent call bot server interlocked with a switch, a TTS server, and a plurality of STT servers of a call center system according to an embodiment of the present invention includes the steps of: (a) receiving a call connection request from a customer through the switch; (b) connecting a call between the exchange and the intelligent call-bot server; (c) converting the customer's voice input from the exchange into a received text through the STT server; (d) analyzing the converted received text to generate an outgoing text; (e) converting the outgoing text into outgoing voice through the TTS server; (f) transmitting the converted outgoing voice to the customer through the exchange; (g) determining a type of STT server capable of improving a speech recognition rate based on the analysis result of the received text; And (h) performing voice recognition by connecting an optimal STT server among the STT servers according to the determined type of the STT server.

The step (g) may determine the type of STT server capable of improving a speech recognition rate by analyzing the received text to determine any one of a speech type, speech topic, and speech intention.

The step (c) may include analyzing the voice data packet of the customer to calculate a silent section or a muted section in which a volume value is less than or equal to a preset reference value; And changing a packet unit before and after the silent period or the attenuated period using a buffer and transmitting the packet to the STT server.

In addition, the unattended counseling method according to an embodiment of the present invention further includes providing link information for confirming the conversation history and voice recording between the customer and the intelligent call bot server to the customer's terminal connected through the network. can do.

In addition, an unattended counseling method according to an embodiment of the present invention includes: transmitting identity authentication request information to the customer's terminal when click information on the link information is received from the customer's terminal; And transmitting the voice recording file to the customer's terminal when the user authentication completion information is received from the customer's terminal.

In addition, the unattended counseling method according to an embodiment of the present invention further includes providing a scenario management screen capable of checking and changing a conversation scenario between the customer and the intelligent call bot server in real time to a counselor terminal connected through a network, , The scenario management screen may be configured to form a flow by connecting a plurality of nodes having different functions in a drag-and-drop method.

The intelligent call bot server and the unmanned counseling method using the same according to an embodiment of the present invention can improve the speed and accuracy of voice recognition to output a real-time conversation history.

In addition, the intelligent call-bot server and the unattended counseling method using the same according to an embodiment of the present invention may provide a conversation history and a voice recording file to a counselor and a customer.

In addition, the intelligent call-bot server and the unmanned counseling method using the same according to an embodiment of the present invention allow a plurality of nodes to be connected in a drag-and-drop method in a web browser to form a flow, so that scenario management can be conveniently performed. do.

The effects according to the present invention are not limited by the contents exemplified above, and more various effects are included in the present specification.

1 is a schematic configuration diagram of a call center system to which an intelligent call bot server according to an embodiment of the present invention is applied.
2 is a schematic configuration diagram of a call center system for explaining the detailed configuration and operation of an intelligent call bot server according to an embodiment of the present invention.
3 is a schematic flowchart illustrating a method for an intelligent call bot server to perform unmanned consultation according to an embodiment of the present invention.
4 is a flowchart illustrating a detailed method of resizing a packet transmission unit in step S330 in FIG. 3.
5 is a flowchart illustrating a detailed method of determining an optimal STT server type in step S360 in FIG. 3.
6 is a flowchart illustrating a method of transmitting a conversation history and recording file to a customer by an intelligent call bot server according to an embodiment of the present invention.
7 is a diagram illustrating a state of utterance of a specific consultation provided by an intelligent call bot server to a counselor terminal according to an embodiment of the present invention.
8 is a diagram illustrating a counseling history of a specific counselor and a detailed counseling status for a specific counseling provided by an intelligent call bot server to a counselor terminal according to an embodiment of the present invention.
9 is a diagram illustrating a scenario management surface provided by an intelligent call bot server to a counselor terminal according to an embodiment of the present invention.

Advantages and features of the present invention, and a method of achieving them will become apparent with reference to the embodiments described below in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but may be implemented in a variety of different forms, and only these embodiments make the disclosure of the present invention complete, and are common knowledge in the technical field to which the present invention pertains. It is provided to completely inform the scope of the invention to those who have, and the invention is only defined by the scope of the claims. The same reference numerals refer to the same elements throughout the specification.

Hereinafter, an intelligent call bot server and an unmanned counseling method using the same according to embodiments of the present invention will be described with reference to the drawings.

1 is a schematic configuration diagram of a call center system to which an intelligent call bot server according to an embodiment of the present invention is applied.

Referring to FIG. 1, it can be seen that the call center system is connected to a data network (network network) and a telephone network (voice network). The solid line indicates the connection through the data network, and the dotted line indicates the connection through the telephone network.

FIG. 1 shows that the intelligent call bot server 100 according to an embodiment of the present invention is installed in connection with an existing customer's call center system. Since the intelligent call-bot server 100 can also serve as an automatic response system called ARS (Audio Response System) or IVR (Interactive Voice Response), it can be interlocked and installed even when there is no IVR as shown in FIG. 1.

Of course, even if there is an IVR in the existing customer's call center system, the intelligent call bot server 100 according to an embodiment of the present invention may be interlocked and installed. In this case, when a call is received from the customer, the call is branched through the customer's IVR, and the callbot server 100 may be operated.

Switchboard (IPPBX, 10) of existing customer system of intelligent call-bot server (100), STT server (20), TTS server (30), computer telephone integration system (CTI, 40), recording server (REC, 50), web application A server (WAS, 70), a database management system (DBMS, 70), and may be installed in conjunction with the electronic display (80).

When a customer makes a call to the call center, they connect to the switchboard (IPPBX, 10) through the telephone network and request a consultation. In this case, the switch 10 is shown as being a VoIP-based IPPBX in FIG. 1, but is not limited thereto and is a concept encompassing a conventional PBX.

The switch 10 connects the call with the intelligent call bot server 100 so that the customer and the call bot server 100 can make a call.

The intelligent call bot server 100 plays a role of unattended consultation with the customer, and can perform various consultations such as after-sales service application, delivery order, civil petition application, insurance contract signing, mobile phone purchase contract, and mobile phone plan change.

The switch 10 and the intelligent call bot server 100 are connected to each other through a telephone network, and may also be connected to the counselor groups 3 and 5 through a telephone network. Here, the counselor groups 3 and 5 may refer to groups classified according to counseling content or other criteria.

The switch 10 serves to branch to the intelligent call bot server 100 when a call for consultation request such as service reception is received from the customer, and the intelligent call bot server 100 plays a role of unattended consultation according to a preset scenario. do.

In addition, switchboard (10), intelligent call-bot server (100), STT server (20), TTS server (30), CTI (40), recording server (50), WAS (60), DBMS (70), and billboard (80) Can be configured by being connected to the internal data network.

The STT server 20 serves to convert voice data into text data, and the TTS server 30 serves to convert text data into voice data. Specifically, when voice data is received from a customer, the STT server 20 converts it into text data and transmits it to the intelligent call bot server 100, and when the intelligent call bot server 100 generates text data to be transmitted to the customer, the TTS server (30) plays a role of receiving this, converting it into voice data, and transmitting it to the customer through the exchange (10).

As the STT server 20, a Long Short Term Memory (LTSM) engine based on an algorithm based on a recurrent neural network (RNN) having strengths in unstructured natural language speech recognition may be used.

As the TTS server 30, a DNN-based natural language speech synthesis engine in which a deep neural network (DNN) method is added to a unit selection speech (USS) method may be used.

The CTI 40 plays a role of efficiently distributing and managing calls to the call center system by combining a computer and a telephone. In the CTI 40, information, such as consultation details, etc., are displayed in real time through a computer screen during a call with a customer, thereby enabling efficient call center operation.

The recording server 50 serves to record and store the voice call between the counselor of the call sensor and the customer.

The WAS 60 is a software management server that manages each application executed in the data network of the call center system, and serves to operate the application at the server side.

The DMBS 70 manages a database built through software for information such as consultation details of customers and counselors in the call center system.

The electronic display 80 serves to output a screen for monitoring the operating status of the entire system of the call center.

In FIG. 1, all configurations of the call center system are shown as being built in an internal data network, but an embodiment in which a specific configuration is built in the form of a cloud service in an external data network may also be possible.

For example, the STT server 20 and the TTS server 30 may be formed in an external data network, and the intelligent call-bot server 100 of the present invention may be installed in the “DMZ Zone” of the internal data network. Here, “DMZ Zone” refers to an area that is located in the internal data network but can be accessed from the external data network.

The customer is connected to the intelligent call bot server 100 through a telephone network through a terminal such as a telephone, and if necessary, information such as text messages and video data can be transmitted and received through the Internet network.

The intelligent call bot server 100 of the present invention can automatically select an optimal server from among a plurality of STT servers 20 in order to improve the speed and accuracy of real-time voice recognition and output of conversation history, and is transmitted to the STT server 20 A function such as resizing a packet unit of voice data to be generated may be performed, which will be described in detail below with reference to FIG. 2 and the like.

2 is a schematic configuration diagram of a call center system for explaining the detailed configuration and operation of an intelligent call bot server according to an embodiment of the present invention.

2, for convenience, the entire configuration of the call center system is not shown, and the intelligent call bot server 100 includes a switch 10, an STT server 20, a TTS server 30, a recording server 50, a customer terminal 2, and a consultant. Only connected to the terminal 4 is shown.

When a call is received from the customer terminal 2, the switch 10 branches the call to the intelligent call-bot server 100 so that a desired consultation can be performed.

The intelligent call bot server 100 includes a chatbot server 110, a call processing unit 120, a consultation processing unit 130, a consultation content storage unit 140, a consultation content output unit 150, an STT server determination unit 160, and a scenario. It may be configured to include the management unit 170.

When a call is connected to the customer terminal 2 through the switch 10, the consultation processing unit 130 converts the customer's voice into a received text through the STT server 20 and transmits it to the chatbot server 110, and the chatbot server 100 ) Receives the outgoing text generated by the TTS server 30, converts it into an outgoing voice through the TTS server 30, and transmits the converted outgoing voice to the switchboard 10 to handle counseling.

The chatbot server 110 analyzes the received text to generate an outgoing text, and plays a role of generating an output comment for each scenario, and a preprocessor 112, a colloquial conversion unit 114, a database 116, and a comment output unit Consists of 118.

The pre-processing unit 112 serves to divide the received text into a unit having meaning and to divide the received text into words.

The colloquial conversion unit 114 serves to refine or replace the received text preprocessed by the preprocessor 112.

The database 116 serves to pre-build and store data that is the basis for the colloquial conversion by the colloquial conversion unit 114.

The comment output unit 118 serves to generate and output an outgoing text and an output comment for each scenario based on the content converted by the colloquial conversion unit 114.

For example, the colloquial conversion unit 114 may manage words to distinguish “positive/negative” and utilize them in a bot conversation scenario. When the database 116 stores “like, right” as a positive expression, and stores “no, not” as a negative expression, and then the customer's voice is converted into input text through the STT server 20 and received. The contents stored in the database 116 may be compared with the contents stored in the database 116, and may be converted by the colloquial conversion unit 114 in such a manner as to substitute a similar word or refine a phrase by a pre-built algorithm.

In addition, the colloquial conversion unit 114 may generate an optimal output text by building data in the database 116 in advance in a similar manner to the above, even in the case of an investigation including a number such as “date” or “amount”.

In addition, the database 116 may store chatbot scenarios and output comments according to counseling categories, questions and answers, and the comment output unit 118 is a customer based on the text converted by the colloquial conversion unit 114. You can print a comment that will be sent to you.

The call processing unit 120 is responsible for managing the conversation history and voice recording between the intelligent call bot server 100 and the customer, and the bot conversation history management unit 122, the recording management unit 124, the authentication processing unit 126 and the operating standards It is configured to include a management unit (128).

The bot conversation history management unit 122 manages the conversation history between the customer and the intelligent call bot server 100 to be tracked based on the customer's phone number and name. At this time, the conversation history refers to text data obtained by converting a customer's voice into text and text data generated by the comment output unit 118.

The recording management unit 124 serves to record the contents of the voice conversation between the customer and the intelligent call bot server 100 through the recording server 50.

At this time, under the control of the recording management unit 124, the recording server 50 may record only the customer's voice or simultaneously record the voice provided by the intelligent call bot server 100 synthesized.

In addition, the recording management unit 124 may record the voices of the customer and the call bot server 100 in a form that can be reproduced by tracking each query item of the intelligent call bot server 100.

In this case, a method of recording in a form that can be traced for each query item by the recording server 50 may use a technique of indexing so as to be traceable for each query item when creating a recording file.

The authentication processing unit 126 transmits authentication request information to the customer terminal 2 when a process of self-authentication is required during the consultation process, and enables a preset next procedure to be performed when authentication success is confirmed.

For example, the consultation content output unit 150, which will be described later, may provide download link information for a conversation history and voice recording file to the customer terminal 2, where the customer When the information is clicked, the authentication processing unit 126 transmits a message for requesting authentication of the user, and when the authentication is successful, the conversation history and the voice recording file may be downloaded.

The identity authentication may be performed directly by the authentication processing unit 126, but may be performed through a communication company server, a bank server, etc. linked to the intelligent call-bot server 100.

The operation standard management unit 128 plays a role of managing the operation time of the intelligent callbot server 100, the time interval, and the standard for connection to a counselor. For example, the operation standard management unit 128 may set an operation time, etc. based on a consultation category, content, and the like.

The consultation contents storage unit 140 serves to store the consultation contents processed by the consultation processing unit 130. For example, if the customer's consultation content is an AS receipt for a product, the AS receipt details can be organized and stored. At this time, the AS received details may include customer information, AS target product, AS required content, counselor information, conversation history information, voice recording history, and the like.

The counseling content output unit 150 serves to output the counseling content stored in the counseling content storage unit 140 to the counselor terminal 4 or the customer terminal 2.

For example, the counseling content output unit 150 may output the counseling situation and content for each counselor to the counselor terminal 4 in real time. In particular, the counseling situation and content that the counseling content output unit 150 outputs in real time may be a manager terminal among the counselor terminals 4.

In addition, if necessary, the counseling content output unit 150 may provide individual counseling content to a specific customer in real time or non-real time. For example, a conversation history may be provided to the customer in the form of text or chat to a customer who conducts a consultation, and a voice recording file may be provided when the consultation is over.

Of course, as described above, the conversation history and the voice recording file may not be provided immediately, but may be provided only when the user authentication is successful.

The STT server determination unit 160 serves to determine and connect an optimal STT server capable of improving the accuracy and speed of speech recognition among the plurality of STT servers 20.

As shown in Figure 2, the STT server 20 may be configured to include STT server #1 (22), STT server #2 (24), and STT server #3 (26). Each STT server may be a server having different accuracy for a specific category of voice.

For example, the STT server 20 may be configured to include an AI-based STT server provided by a company such as Google and Watson, and a professional-based STT server provided by a company such as Lead Speaker Korea and Solugate. .

The STT server determination unit 160 may determine the optimal STT server type based on the text information converted by the preprocessing unit 112 and the colloquial conversion unit 114 of the chatbot server 110 or a recognized consultation category. It will be described later with reference to FIG. 5.

The packet processing unit 170 plays a role of resizing the size of a customer's voice data packet transmitted through the switch 10.

In general, voice data is transmitted in the form of an RTP packet, and the RTP packet is generally transmitted in a fixed packet size such as 320 bytes and 640 bytes. However, in this case, the speed and accuracy of speech recognition of the STT server 20 are degraded due to interruptions in the words of the sentence or the like.

Accordingly, the packet processing unit 170 analyzes the packet, detects a silent period or a mute period in which the volume value is less than a preset reference value, divides and resizes the packet before and after the silent period or the mute period, and transmits the packet to the STT server 20.

A detailed resizing method of the packet processing unit 170 will be described later with reference to FIG. 4.

The scenario management unit 180 serves to provide a scenario management screen for real-time confirmation and change of a conversation scenario between a customer and the intelligent call bot server 100 to the counselor terminal 4 connected through a network.

In particular, the scenario management unit 180 may provide a web-based scenario management screen for forming a flow by connecting a plurality of nodes having different functions in a drag-and-drop manner.

3 is a schematic flowchart illustrating a method for an intelligent call bot server to perform unmanned consultation according to an embodiment of the present invention.

When a call between the customer terminal 2 and the intelligent call bot server 100 is connected, the intelligent call bot server 100 transmits a counseling guide (S310).

Next, the intelligent call bot server 100 transmits a consultation query according to a pre-built scenario (S315).

When the customer's response voice information is received (S320), the packet processing unit 170 resizes the packet transmission unit (S330).

At this time, the packet processing unit 170 resizing the packet transmission unit is to improve the speed and accuracy of speech recognition of the STT server 20, and changes the packet size based on the silence or attenuation interval of the voice data. It will be described later with reference to FIG. 4.

When the resized packet is delivered to the STT server, the intelligent call-bot server 100 converts the received text into a received text through the STT server 20 (S340).

Although not shown in FIG. 3, the received text may be converted into highly accurate data through the preprocessor 112 and the colloquial conversion unit 114.

Next, based on the converted text, it is determined whether an answer suitable for the answer criterion has been received for the query item (S350).

If it is determined that the answer criteria are appropriate in step S350, the STT server determination unit 160 determines the optimal STT server type (S360), connects to the optimal STT server (S365), and returns to step S315 to query the next item. Send out a comment.

If it is determined in step S350 that the answer criteria are not met, it is determined whether the number of failures to answer is greater than a preset reference value (S370).

If it is determined in step S370 that the number of failures to answer is not greater than the reference value, the process returns to step S315 and transmits a material comment for the same item.

If it is determined that the number of failures to answer is greater than the reference value in step S370, it is processed as a counseling failure (S375).

In this way, the counseling query is sent and the answers are inputted. When all answers to all the query items are received, the counseling ends and the counseling contents are saved.

The embodiment of FIG. 3 shows a method of selecting and connecting an optimal STT server by analyzing text information that is a result of speech recognition of a customer, and determining an optimal STT server type based on the corresponding result. In this method, the STT server for the first customer's voice is set as a default value, and the optimal STT server is selected and connected as a result of the initial voice recognition.

However, in addition to this method, if items such as a consultation category selected by the customer at the start of the consultation are preset, the STT server that performs speech recognition for the first customer's voice according to the category is selected as the optimal STT server to perform speech recognition How to do it would be possible.

In this case, the optimal STT server type based on the consultation category, etc., has been established as a database in advance, and an initial STT server can be automatically selected and connected based on the database.

4 is a flowchart illustrating a detailed method of resizing a packet transmission unit in step S330 in FIG. 3.

First, a customer's voice information packet is received (S410). In this case, the voice information may be an RTP packet. The intelligent call-bot server 100 analyzes the voice information packet (S415), and determines whether the corresponding packet is silent or has a volume value smaller than a preset reference value (S420).

If it is determined in step S420 that the corresponding packet has a silence or a volume value less than a preset reference value, the number of silences is counted (S430), and the current packet is copied to the previous buffer (S440).

In step S420, if it is determined that the packet is not silent or the volume value is less than a preset reference value, it is determined whether the previous buffer is empty (S421).

If the previous buffer is not empty in step S412, the previous buffer is stored in the new buffer (S423).

If it is determined that the previous buffer is empty after step S423 or in step S412, the packet is stored in the new buffer (S425).

The next silence count is initialized (S427), and the previous buffer is initialized (S429).

After step S440 or after step S429, it is determined whether the number of silences is greater than a preset reference value (S450).

If it is determined in step S450 that the number of silences is not greater than the reference value, the process returns to step S410 to receive the next packet.

If it is determined in step S450 that the number of silences is greater than the reference value, the silence is stored in front of the new buffer (S470), and the new buffer is transmitted to the STT server 20 (S470).

The next silent number is initialized (S480), a new buffer is also initialized (S490), and the process returns to step S410 to receive the next packet.

In the same way as described above, if the packet is silent or the volume value is attenuated below a certain value, the packet is reclassified based on the number of silences equal to the preset number, readjusted the size, and then transmitted to the STT server 20 to speed up speech recognition. And accuracy can be improved.

5 is a flowchart illustrating a detailed method of determining an optimal STT server type in step S360 in FIG. 3.

The intelligent call-bot server 100 analyzes the received text (S510) and determines whether a speech subject is identified (S520).

When the speech topic is identified, STT server type information corresponding to the speech topic is called and selected (S570).

If the subject of speech is not identified in step S520, it is determined whether the speech form is a natural language form (S530).

If it is determined that the speech form is a natural language form in step S530, it is determined as an AI-based STT server (S550).

If it is determined that the utterance form is not a natural language form in step S530, it is determined whether there are a plurality of utterance intentions (S540).

If it is determined that there are a plurality of utterance intentions in step S540, it is determined as an AI-based STT server (S550).

At this time, the AT series STT server may be an STT server such as Google or Watson.

If it is determined that the utterance intention is singular in step S540, it is determined as a specialized STT server (S560).

At this time, the STT server of a specialized series may be an STT server such as Lead Speaker Korea or Solugate.

Fig. 5 is only a method according to an embodiment, and it may be possible to proceed with another algorithm.

For example, the identification of the speech topic in step S520 may be performed not by text analysis, but by category information input by the user at the beginning of the consultation as described above.

In addition, in step S530, if the speech form is not in the natural language form, the process may proceed directly to step S560 and may be determined as a specialized STT server.

In addition, step S540 may be used as an independent condition independently of step S530 rather than after step S530.

6 is a flowchart illustrating a method of transmitting a conversation history and recording file to a customer by an intelligent call bot server according to an embodiment of the present invention.

The intelligent call-bot server 100 of the present invention may transmit the conversation history and download link information for the recording file to the customer terminal 2 (S610).

When the customer clicks the link information on the customer terminal 2, the intelligent call-bot server 100 receives the link click completion information of the customer (S620).

The intelligent call-bot server 100 may transmit personal authentication request information to the customer terminal 2 (S630). Of course, at this time, the intelligent callbot server 100 may transmit the authentication request message through an external interworking server instead of directly transmitting the authentication request information.

Next, the intelligent call bot server 100 determines whether or not the identity authentication is successful (S640).

If authentication is successful, the conversation history and recording file are transmitted (S650), and if not, the authentication failure message is transmitted without transmission, and the transmission is terminated without transmission.

In FIG. 6, although an embodiment in which both the conversation history and the recording file can be downloaded after authentication has been described, it may be possible to provide only one of the two files.

In addition, the conversation history and recording files can be provided to the counselor terminal 4 as well, and in this case, it will be generally configured so that inquiries can be performed by the administrator terminal without the need for personal authentication.

Providing the conversation history and voice recording file to the customer terminal 2 through identity authentication in the same manner as described above will be very effective in terms of providing data for proof of the contract when an electronic contract is made using a call bot.

In this case, when the customer makes an electronic contract while performing communication through the phone network through the customer terminal (2), the contract can be completed only through the phone network, and the download link information as described above is in the form of a text message through the phone network or the Internet network. Will be able to provide.

In the future, identification and download by clicking the link may be performed through the Internet network. In this way, even customers located in island areas where the Internet network is not installed can easily complete an electronic contract through the telephone network, and proof of the contract by downloading conversation history and recording files through self-authentication at a location with an Internet network in the future. Can be facilitated.

7 is a diagram illustrating a state of utterance of a specific consultation provided by an intelligent call bot server to a counselor terminal according to an embodiment of the present invention.

In FIG. 7, the content of consultation between the customer and the ARS is divided and displayed for each conversation. Specifically, the calling number, date and time, speaker, result, score, and audio items are displayed. The utterances of the customer and ARS are displayed in text in the ‘Result’ column, the ‘Score’ column shows the customer’s response as a score, and in the ‘Audio’ column, you can play the recorded file of the corresponding utterance. Play icon is displayed.

8 is a diagram illustrating a counseling history of a specific counselor and a detailed counseling status for a specific counseling provided by an intelligent call bot server to a counselor terminal according to an embodiment of the present invention.

In FIG. 8, it can be seen that the counseling history for the month of April performed by the counselor of number '1007' is displayed, and detailed counseling contents of the selected specific counseling are displayed on the right side.

Information on'CALL ID','start date and time','end date','recording time', and'STT engine' for each consultation is displayed. Detailed conversations between customers are displayed.

The counseling status information as shown in FIGS. 7 and 8 can be provided in real time. As described above, the intelligent call bot server 100 of the present invention resizes the packet size and transmits it to the STT server 20, thereby improving the accuracy of speech recognition. By automatically selecting the optimal STT server 20 that can be improved, it is possible to improve speed and accuracy in providing real-time conversation history information as described above.

9 is a diagram illustrating a scenario management surface provided by an intelligent call bot server to a counselor terminal according to an embodiment of the present invention.

As shown in the screen of FIG. 9, the scenario can be changed by moving a plurality of nodes having different functions arranged on the left side of each scenario in a drag-and-drop method to form a connection relationship with each node.

In addition, when a specific node is selected, a screen for inputting detailed attribute values for the corresponding node is displayed on the far right, allowing the administrator to input setting values for each node.

As shown in FIG. 9, by adopting a method of manipulating nodes for each function in a web-based manner in a drag-and-drop method, it is possible to easily grasp the flow of complex scenarios and easily change the scenario.

As described above, an intelligent call bot server and an unattended reception method using the same according to embodiments of the present invention have been described.

Those of ordinary skill in the art to which the present invention pertains will appreciate that the present invention can be implemented in other specific forms without changing the technical spirit or essential features thereof. Therefore, it should be understood that the embodiments described above are illustrative in all respects and are not limiting. The scope of the present invention is indicated by the claims to be described later rather than the detailed description, and all changes or modified forms derived from the meaning and scope of the claims and their equivalent concepts should be construed as being included in the scope of the present invention. do.

2: customer terminal 3, 5: consultant group
4: Consultant terminal 10: Switchboard (IPPBX)
20: STT server 30: TTS server
40: Computer Telephony Integration System (CTI) 50: Recording Server
60: Web Application Server (WAS) 70: Database Management System (DBMS)
80: electronic board 100: intelligent call bot server 100
112: preprocessing unit 114: colloquial conversion unit
116: database 118: comment output
120: call processing unit 122: bot conversation history management unit
124: recording management unit 126: authentication processing unit
128: operation standard management unit 130: consultation processing unit
140: consultation contents storage unit 150: consultation contents output unit
160: STT server determination unit 170: packet processing unit
180: Scenario Management Department

Claims (16)

In the intelligent call bot server that provides unmanned consultation service to customers by interlocking with the exchange of the call center system, TTS server, and multiple STT servers,
The intelligent call-bot server connects a call to the exchange when a customer's consultation request is received through the exchange, converts the customer's voice input to the exchange into a received text through the STT server, and converts the received text Is configured to generate an outgoing text by analyzing the outgoing text, convert the outgoing text into an outgoing voice through the TTS server, and transmit the outgoing voice to a customer through the exchange,
STT to determine the type of STT server capable of improving the speech recognition rate based on the analysis result of the received text, and to perform speech recognition by connecting the optimal STT server among the STT servers according to the determined type of STT server Intelligent call-bot server including a server decision unit. The method of claim 1,
The STT server determination unit analyzes the received text to determine any one of a speech type, speech topic, and speech intention to determine a type of STT server capable of improving a speech recognition rate. The method of claim 2,
The STT server determination unit has previously built an STT type database on type information of the optimal STT server for each utterance topic, and selects and connects the STT server optimal for the utterance topic based on the STT type database. Intelligent callbot server. The method of claim 2,
The STT server determination unit, when the speech type is a natural language type or when it is determined that the speech intention includes a plurality of intentions, the intelligent call bot server, characterized in that for selecting and connecting an AI-based STT server. The method of claim 2,
The STT server determination unit selects and connects a specialized STT server when the utterance type is a specialized utterance form including a terminology or when the utterance intention is determined to include a single intention. Callbot server. The method of claim 2,
By analyzing the voice data packet of the customer, a silent section or a soft tone section in which the volume value is less than a preset reference value is calculated, and the packet unit is changed before and after the silent section or the attenuated section using a buffer to the STT server. Intelligent call-bot server further comprising a packet processing unit to transmit. The method of claim 2,
Further comprising a call processing unit for managing the conversation history and voice recording between the customer and the intelligent call bot server,
The call processing unit provides the conversation history to a counselor terminal connected through a network in real time, and provides link information for confirming the voice recording. The method of claim 7,
The call processing unit is configured to provide link information for confirming the conversation history and the voice recording to the customer's terminal connected through a network. The method of claim 8,
The call processing unit intelligent call bot server, characterized in that to provide link information for confirming the voice recording when the customer's identity authentication is completed. The method of claim 2,
A scenario management unit that provides a scenario management screen capable of checking and changing a conversation scenario between the customer and the intelligent call bot server in real time to a counselor terminal connected through a network,
The scenario management screen is configured to form a flow by connecting a plurality of nodes each having a different function in a drag-and-drop method. In the unmanned consultation method using an intelligent call bot server interlocked with an exchange of a call center system, a TTS server, and a plurality of STT servers,
(a) receiving a call connection request from a customer through an exchange;
(b) connecting a call between the exchange and the intelligent call-bot server;
(c) converting the customer's voice input from the exchange into a received text through the STT server;
(d) analyzing the converted received text to generate an outgoing text;
(e) converting the outgoing text into outgoing voice through the TTS server;
(f) transmitting the converted outgoing voice to the customer through the exchange;
(g) determining a type of STT server capable of improving a speech recognition rate based on the analysis result of the received text; And
(h) Unmanned counseling method comprising the step of performing voice recognition by connecting an optimal STT server among the STT servers according to the determined type of STT server. The method of claim 11,
In the step (g), the received text is analyzed to determine any one of a speech form, speech topic, and speech intention to determine a type of STT server capable of improving a speech recognition rate. The method of claim 12,
The step (c),
Analyzing the voice data packet of the customer to calculate a silent section or a soft tone section in which a volume value is less than or equal to a preset reference value; And
And transmitting to the STT server by varying a packet unit before and after the silent period or the attenuated period using a buffer. delete delete delete

Images