RU2781493C1 - Robotic question-answer system with a voice module "virtual assistant" - Google Patents

Abstract

FIELD: robotics.

SUBSTANCE: invention relates to interfaces for controlling telecommunication information systems. The technical result is achieved due to the system containing a telephone station module configured to redirect the received call to a module with the bidirectional sound transmission protocol AudioSocket, configured to intercept and redirect voice flow control through the silence detection module into the speech recognition module configured to convert oral speech into text and connected with a dialogue module connected with the reporting module via the virtual assistant database storing the information of the dialogue script and information on passing the script for each request; the dialogue module is also connected with the speech synthesis module connected, in turn, with the module of the bidirectional sound transmission protocol AudioSocket.

EFFECT: ensured speech recognition and interaction with the text communication channel.

1 cl, 1 dwg

Description

The technical solution relates to the field of information technology and communication systems, using, in particular, the interface for managing telecommunication information systems using IVR (Interactive Voice Response - voice menus), and can be used to automatically provide a variety of services in a telecommunications network using the IVR service .

A system is known from the prior art (patent RU 95206, IPC H04L 12/00, published on 06/10/2010), containing the functionality of playing audio recordings in the voice channel with the ability to partially or completely close incoming calls coming to the contact center in the voice channel. Also containing a multi-level voice menu system, including an IVR (Interactive Voice Response - interactive voice response) module, which is connected to a message server, including means for receiving, processing and storing voice messages, means for creating a menu and message queue, a switching module associated with IVR module.

The disadvantages of this system: interaction with the system is carried out by pressing the keys on the phone, depending on the voiced information, as well as the inability to integrate with a text communication channel such as a chat system or messenger.

A robotic question-answer system with a voice module is known from the prior art (patent RU 2724600, IPC G10L15 / 22, G06F40 / 20, publ. 06/25/2020), which contains a module for automatically starting a connection with a user's electronic device, configured to interact with an IVR - a system for providing connection with the user's electronic device, and the system contains a module for reporting on the interaction of the robotic question-answer system with the IVR system, configured to fill the report with data obtained during the interaction of the robotic question-answer system with the IVR system in the absence of established connection of the system with the user's electronic device.

Disadvantages of this system: lack of a visual dialog interface for editing the dialog script.

The technical problem of the invention is aimed at expanding the functionality of the voice robotic question-answer system.

The technical result, to which the invention is directed, is to provide recognition of oral speech, and, due to keywords or a classifier of topics, reproduce a certain block of speech phrases from a dialogue scenario, as well as to ensure interaction with a text communication channel according to the principle of operation of a voice channel, with the exception of the module speech recognition.

The technical result is achieved by the fact that the robotic question-answer system with a voice module contains an Asterisk PBX telephone exchange module, configured to redirect it to a module with an AudioSocket bidirectional audio protocol when a call arrives, configured to intercept control of the voice stream and redirect it through the module silence detection into a speech recognition module configured to convert oral speech into text and connected to the dialogue module, which is connected to the reporting module through the virtual assistant database, which stores information about the dialogue script and information on the passage of the script for each call, also, the dialog module is connected to the speech synthesis module, which in turn is connected to the AudioSocket bidirectional audio protocol module.

Due to the presence of a speech recognition module in the system, the system can interact with the client in the voice channel.

Due to the presence of a speech synthesis module in the system, the system reproduces a certain block of speech phrases without using pre-recorded audio files.

With the help of the dialog module, interactions with the text communication channel are provided according to the principle of the voice channel, with the exception of the speech recognition module, by working with the textual representation of data using the methods of passing the script by keywords or the classifier of topics.

The voice robotic question-and-answer system can be represented by a software and hardware system capable of interacting with the user's electronic device via a digital communication channel, for example, a server containing specialized software and configured for question-answer interaction with the user's electronic device, as well as loading, editing and removal of specialized software by the system administrator. The system may comprise a database of user interaction scripts that may be integrated into the auto-start module for connecting to the user's electronic device, or it may be hosted remotely and connected to the auto-script launch module by any known means of communication.

The essence of the proposed system is illustrated by graphic material, where in Fig. a block diagram of the system is presented, where the following positions are assigned to the main blocks / modules:

1. Phone, the device from which the signal is received;

2. ATC Asterisk telephone exchange module;

3. Module with AudioSocket bidirectional audio protocol;

4. Silence detection module;

5. Speech recognition module;

6. Dialogue module;

7. Virtual assistant database;

8. Reporting module;

9. Speech synthesis module.

The system works as follows.

When implementing the system operation, the following main stages are carried out (shown in Fig.):

10 - Making a call (call) from the user's phone 1 to the PBX Asterisk 2 module;

11 - Transfer control of the voice flow of the PBX Asterisk 2 module with AudioSocket 3;

12 - Extracting parts of the voice stream, where the user spoke any phrases, to the module with AudioSocket 3, using the VAD 4 silence detection module;

13 - Converting voice to text from a phrase extracted thanks to the VAD 4 silence detection module using the speech recognition module 5;

14 - Transferring a text phrase from the speech recognition module 5 to the dialog module 6;

15 - Saving the necessary information to the database 7 from the dialog module 6, to form a response to the user;

16 - Formation of a dialogue using the necessary information from the database 7 processed by the reporting module 8;

17 - Saving the current state of the dialogue with the user to the database 7 from the reporting module 8;

18 - Extracting a text response to the user in the dialog module 6 from the database 7;

19 - Converting a text phrase from a dialogue module 6 into an audio segment using a speech synthesis module 9;

20 - Adding an audio phrase from the speech synthesis module 9 to the voice stream of the module with AudioSocket 3;

21 - Transmission of a voice stream from the module with AudioSocket 3 to the module of the telephone exchange PBX Asterisk 2;

22 - Playback of the voice stream in the user's phone 1 from the PBX Asterisk 2 module.

An example of the operation of the system with incoming calls in the voice channel regarding non-working television.

The system receives a call from the user's device to module 2 of the telephone exchange Asterisk PBX. Then the voice stream is transferred from module 2 of Asterisk PBX to module 3 with AudioSocket. Next, a phrase is expected from the user with the reason for the request. The user voices the reason "does not show cable TV." With the help of the silence detection module 4, the moment of the end of the phrase is determined, after which the silence detection module 4 sends a record of the user's phrase to the speech recognition module 5. The speech recognition model 5 translates it into text and sends it to the dialog module 6. The dialog module 6 saves the recognized text of the user's phrase to the database 7 of the virtual assistant and determines the topic of the appeal using the topic classifier and takes from the database 7 the data that needs to be voiced within the scenario for a specific topic.

In the case of the topic “problems with cable TV”, a phrase will be heard asking the user to check the tightness of the cable connection to the TV. The response is sent to the voice channel from the dialog module 6 to the audio socket module 3 via the speech synthesis module 9, which converts the response text into audio. Further, from module 3 with AudioSocket, the voice stream is transmitted to module 2 of the Asterisk PBX and played back to the user's device. Next, a response from the user is expected.

Subsequent processing of all responses from the user occurs according to a similar scenario for processing the first phrase. In the current example, if the user answers that he has a cable tightly inserted into the TV, the system will prompt him to check the integrity of the cable in the apartment. In the event that the user confirms the integrity of the cable in the apartment, then information is announced that an engineer will visit the address of the user to check the integrity of the cable in the adjacent territory and the conversation will end. Asterisk PBX module 2 will end the call.

The presented system provides automatic informing of clients without the participation of an operator on incoming and outgoing calls, chats, as well as the execution of automation scripts.

This system can be used in chat bots, or voice robots, helps to process incoming customer requests without the help of operators, and allows you to automatically inform customers, automatically close the call, as well as automatically determine the subject of the call and transfer the call to the right employee of the company .

Claims (1)

A robotic question-answer system with a voice module, characterized in that it contains an Asterisk PBX telephone exchange module, configured to redirect it to a module with an AudioSocket bidirectional audio protocol when a call arrives, configured to intercept control of the voice stream and redirect it through the detection module silence into a speech recognition module configured to convert oral speech into text and connected to the dialogue module, which, through the database of the virtual assistant, which stores information about the dialogue script and information on the passage of the script for each call, is connected to the reporting module, also the dialog module is connected to the speech synthesis module, which in turn is connected to the AudioSocket bidirectional audio protocol module.

Images