JP2014178381A - Voice recognition device, voice recognition system and voice recognition method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a voice recognition device, a voice recognition system and a voice recognition method capable of efficiently improving a voice recognition rate.SOLUTION: The voice recognition device includes a dictionary management information storage part, a dictionary storage part, a voice storage part, a call information storage part, a dictionary selection part and a voice recognition part. The dictionary storage part stores a voice recognition dictionary per operator and a voice recognition dictionary per job type. The voice storage part stores voice of each operator who has performed telephone correspondence to a user. The call information storage part stores call information associating identification information of each operator who performed telephone correspondence to a user with identification information of the type of a job that he/she is in charge of. The dictionary selection part selects a dictionary to be used for voice recognition by referring to the dictionary management information storage part on the basis of the call information. The voice recognition part converts voice of an operator read from the voice storage part into text data by referring to the selected dictionary of the dictionary storage part.

Description

  Embodiments described herein relate generally to a speech recognition apparatus, a speech recognition system, and a speech recognition method.

  In recent years, there are various business forms such as manufacturing, sales, and finance, but there are cases where a call center is set up where customers and operators answer the phone as part of customer service. Many.

  In the call center, a voice recognition device is used to record the call contents between the customer and the operator and convert the recorded voice data into text in order to check the quality of the operator's response to the customer and to use the voice of the customer. It has been.

  In order to confirm the quality of the operator's response to the customer, supervisors and other managers can display the character strings that have been recorded on the voice recognition device and converted to text on the monitor of the monitoring terminal. May search for keywords such as inappropriate terms, or search for words that are necessary to understand customer needs.

  Incidentally, a dictionary for speech recognition is used in the speech recognition apparatus. Usually, such a dictionary is a general-purpose dictionary.

  In a call center, the number of operators who deal with customers is not limited to one, but a plurality of people often take turns, and each operator has a different utterance method. Also, call centers often divide their work according to the content of inquiries from customers, and there are many cases where specialized operators respond to each work.

JP 2011-141349 A

  However, in the case of conventional speech recognition devices, since a general-purpose dictionary is used, there is a problem that the speech recognition rate is not good due to the influence of the operator's utterance method and the misrecognition of technical terms used in business. It was.

  The problem to be solved by the present invention is to provide a voice recognition device, a voice recognition system, and a voice recognition method capable of improving the voice recognition rate efficiently.

  The speech recognition apparatus according to the embodiment includes a dictionary management information storage unit, a dictionary storage unit, a speech storage unit, a call information generation unit, a dictionary selection unit, and a speech recognition unit. In the dictionary management information storage unit, operator identification information corresponding to a telephone call to a user, business type identification information indicating a user's inquiry destination, and dictionary identification information used for voice recognition are stored in association with each other. Yes. The dictionary storage unit stores a speech recognition dictionary for each operator and a speech recognition dictionary for each business type. The voice storage unit stores an operator voice corresponding to the user. The call information generation unit stores call information in which the identification information of the operator who corresponds to the user by telephone and the identification information of the business type handled by the operator are associated with each other. The dictionary selection unit selects a dictionary to be used for speech recognition with reference to the dictionary management information storage unit based on call information. The voice recognition unit refers to the dictionary in the dictionary storage unit selected by the dictionary selection unit, and converts the operator's voice read from the voice storage unit into text data.

It is a figure showing the outline composition of the voice recognition system of an embodiment. It is a block diagram which shows the apparatus structure by the side of a call center. It is a figure which shows one example of dictionary management DB. It is a flowchart which shows operation | movement of a speech recognition system.

Hereinafter, embodiments will be described in detail with reference to the drawings.
(Embodiment) FIG. 1 is a diagram showing the configuration of a speech recognition system according to an embodiment.

  As shown in FIG. 1 and FIG. 2, the voice recognition system of this embodiment includes a user's telephone 1, a public switched telephone network 2: Public switched telephone networks 2 (hereinafter referred to as “PSTN 2”, IP-PBX / PBX, etc.). 3. Operator telephone 4 and computer 5 as an operator operation terminal (hereinafter referred to as “operator PC 5”), control device 6, Computer Telephony Integration device 7 (hereinafter referred to as “CTI device 7”), monitoring terminal Monitoring PC 8, call recording device 9, and speech recognition device 10.

  The circuit switching device 3 performs incoming call control and outgoing call control from the PSTN 2 at the call center. The operator telephone 4 is a telephone for an operator to receive a telephone call from a customer (also referred to as a customer or a user) and perform a telephone service.

  The operator PC 5 is an information terminal for displaying information on the user from the control device 6 and information input from the keyboard by the operator on the screen so that the operator can perform a telephone service.

  The control device 6 controls the CTI device 7, the call recording device 9, the voice recognition device 10, and the like, assigns information, gives instructions to each device, and the like. The control device 6 has an automatic answering function for the user's telephone, prompts the user's telephone operation, specifies an inquiry destination (business) according to the telephone operation, and connects the user's telephone to the operator telephone 4 for the specified business. Let

  The automatic response function supports identification of business type (business ID) according to a preset response message and the number (tone) of the telephone button associated with the response message, and supports the business of this business ID Operator identification information (operator ID), telephone identification information, and the like are set, the business ID is specified by the telephone button operation (tone transmission operation) performed by the user, and the user's incoming call is currently free This is a function for connecting to the operator telephone 4.

  The CTI device 7 receives from the control device 6 information (such as the user's telephone number and business ID) that the operator supports via the line switching device 3, and generates call information (a pair of operator ID and business ID) (see FIG. 2), and stored in the storage DB 12.

  In other words, the CTI device 7 is a call information generating unit that generates call information in which an operator ID corresponding to a telephone call to a user is associated with a business ID of a business type handled by the operator.

  The call recording device 9 records (stores) a conversation (call contents) between the user and the operator as a separate audio file by associating (corresponding) with an incoming call ID (not shown). That is, the call recording device 9 records the operator's voice corresponding to the user.

  The monitoring PC 8 is a terminal used by an administrator such as a supervisor. The monitoring PC 8 causes the voice recognition device 10 to recognize the content of the call recorded by the call recording device 9 and display the character string converted into text on the monitor, and the administrator uses words that are inappropriate as words spoken by the operator. Search for a keyword such as a certain NG word, or search for a word that is necessary to understand the customer's needs.

  The call recording device 9 divides the contents of the call during the telephone correspondence business into an operator's voice file and a customer's voice file, that is, stores the voice file in the storage DB 12 of the voice recognition device 10.

  As shown in FIG. 2, the speech recognition apparatus 10 includes a dictionary management data base 11 (hereinafter referred to as a database), a storage DB 12, a dictionary selection unit 13, a plurality of dictionaries 14 for each operator, and a plurality of dictionaries 15 for each business. Are stored (stored), a dictionary storage unit 16, a speech recognition engine 17, and a speech recognition result DB 18.

  As shown in FIG. 3, the dictionary management DB 11 stores an operator ID, a business ID, and a dictionary ID that is an identifier of a dictionary to be used in association with each other. The dictionary ID includes an operator dictionary ID and a business dictionary ID, and there may be two cases or one case.

  That is, the dictionary management DB 11 has an operator ID (operator identification information) corresponding to a telephone call to a user, a business ID (business type identification information indicating a user inquiry destination), and a dictionary ID (identification of a dictionary used for voice recognition). Information management) is stored in association with the dictionary management information storage unit.

  In the storage DB 12, the call contents (operator's voice file and customer's voice file) recorded by the call recording device 9 are associated with an incoming call ID (not shown) and stored (stored) as separate voice data. Is done. Further, the storage DB 12 stores call information in which an operator ID corresponding to a telephone call to a user is associated with a business ID of a business type handled by the operator.

  The dictionary selection unit 13 refers to the dictionary management DB 11 based on the call information (a pair of operator ID and business ID) in the storage DB 12 generated by the CTI device 7, and selects a dictionary ID to be used for voice recognition. The selected dictionary ID is output (notified) to the speech recognition engine 17.

  The dictionary storage unit 16 stores a voice recognition dictionary 14 for each operator, a voice recognition dictionary 15 for each business, and a general-purpose dictionary (not shown). The dictionary 14 is a dedicated dictionary in which text data and voice data are registered in association with each operator's pronunciation and speaking method (speaking method). The dictionary 15 is a dictionary in which text data and audio data of technical terms for each business type are linked and registered.

  The speech recognition engine 17 reads out a speech file to be speech-recognized, for example, an operator's speech file from the storage DB 12, and uses the dictionary ID in the dictionary storage unit 16 using the dictionary ID notified (input) from the dictionary selection unit 13. 15, the voice recognition processing of the operator's voice file read from the storage DB 12 is performed, and the voice recognition result is stored in the voice recognition result DB 18.

  In the case where voice recognition is performed in real time on the state of a call between an operator and a user, the following is performed. That is, the voice recognition engine 17 refers to the dictionaries 14 and 15 of the dictionary storage unit 16 selected by the dictionary selection unit 13 and converts the operator's voice recorded by the call recording device 9 into text data.

  In the speech recognition result DB 18, text data and a speech file from which the text data is recognized are associated and stored as a speech recognition result of the speech recognition engine 17.

  That is, the speech recognition engine 17 reads the speech recognition target speech file stored in the storage DB 12 based on the dictionary ID selected from the plurality of dictionaries by the dictionary selection unit 13 and matches the notified dictionary ID. Speech recognition processing is performed with reference to the dictionaries 14 and 15 in the dictionary storage unit 16.

The operation of the speech recognition system of this embodiment will be described below with reference to the flowchart of FIG. FIG. 4 is a flowchart showing the operation of the voice recognition system.
In the case of this embodiment, when a call is made from the user to the representative number of the telephone support desk at the call center, the call is received to the circuit switching device 3 through the PSTN 2 and information on the incoming call (user's phone number) is sent to the control device 6. It is done.

  The control device 6 sends a response message indicating whether the user is already registered or what kind of inquiry is made, and selects the inquiry destination (business) on the center side to the user by operating the numeric keypad of the telephone 1 Let

  When the user is a registered user and the business ID of the inquiry destination is specified by the user's selection operation, the control device 6 calls the telephone 4 of the operator of the business corresponding to the circuit switching device 3 from the user. Instruct to connect (incoming call).

  The circuit switching device 3 connects the telephone (incoming call) from the user to the operator telephone 4 instructed by the control device 6.

  The control device 6 instructs the call recording device 9 to record the call from the time when the operator responds with the telephone 4, whereby the call recording device 9 starts the call recording and links the content of the call to the incoming call ID. An attached audio file is generated (step S101 in FIG. 4) and stored in the storage DB 12.

  The CTI device 7 receives call information (incoming call ID, operator ID, business ID, user telephone number, etc.) from the control device 6 and starts call communication between the operator and the user. A pair of operator ID and business ID) is generated (step S102) and stored in the storage DB 12. As a result, the call information (operator ID, business ID) associated with the incoming call ID and the audio file are stored in the storage DB 12 almost simultaneously.

  The dictionary selection unit 13 reads the call information (operator ID, business ID) stored in the storage DB 12, and uses the call information (operator ID, business ID) as a key to refer to the dictionary management DB 16 to call information (operator ID). , A dictionary ID for speech recognition that matches the business ID) is selected (selected) (step S103), and is output (notified) to the speech recognition engine 17.

  If there is no dictionary ID matching the call information (operator ID, business ID) in the dictionary management DB 11, the dictionary selection unit 13 notifies the speech recognition engine 17 of the dictionary used for speech recognition as a general-purpose dictionary. Matching with call information (operator ID, business ID) may be either operator ID or business ID.

  The speech recognition engine 17 searches the dictionary storage unit 16 using the notified dictionary ID as a key, refers to the corresponding dictionary, and recognizes the speech file to be speech recognized (step S104).

  In the speech recognition apparatus 10, the text file resulting from the recognition processing performed by the speech recognition engine 17 is output to the speech recognition result DB 18 (step S105), and the text file and the speech file are associated with each other and stored in the speech recognition result DB. .

  The monitoring PC outputs the contents of the conversation between the user and the operator in real time, or accesses the voice recognition result DB 18 to read out desired data and display and / or reproduce the voice.

  As described above, according to the first embodiment, a plurality of voice recognition dictionaries 14 and 15 are provided for each operator / work, and used based on call information (operator ID and work ID) generated by the CTI device 7. The dictionary to be selected is selected from a plurality of dictionaries 14 and 15, and voice recognition is performed with reference to the selected dictionary. Therefore, the operator's voice can be recognized by correctly recognizing differences in the utterance methods such as voice peculiar to the operator and intonation. The recognition rate of terms used in business can be improved by using the technical term dictionary 15 for each business. As a result, the speech recognition system as a whole can be efficiently improved. Can be improved.

  Although the embodiment of the present invention has been described, this embodiment is presented as an example and is not intended to limit the scope of the invention. The novel embodiment can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the scope of the invention. This embodiment and its modifications are included in the scope and gist of the invention, and are included in the invention described in the claims and the equivalents thereof.

  Each component shown in the above embodiment may be realized by a program installed in a storage such as a hard disk device of a computer, and the program is stored in a computer-readable electronic medium: program. The computer may realize the functions of the present invention by causing the computer to read from the electronic medium. Examples of the electronic medium include a recording medium such as a CD-ROM, flash memory, and removable media. Further, the configuration may be realized by distributing and storing components in different computers connected via a network, and communicating between computers in which the components are functioning.

  DESCRIPTION OF SYMBOLS 1 ... Telephone, 2 ... Public switched telephone network (PSTN), 3 ... Circuit switching apparatus (IP-PBX / PBX), 4 ... Operator telephone, 5 ... Computer (operator PC), 6 ... Control apparatus, 9 ... Call Recording device, 10 ... Voice recognition device, 11 ... Dictionary management database (dictionary management DB), 13 ... Dictionary selection unit, 14 ... Dictionary for each operator, 15 ... Dictionary for each operation, 16 ... Dictionary storage unit, 17 ... Voice Recognition engine.

Claims (7)

A dictionary management information storage unit that stores the identification information of the operator corresponding to the user, the identification information of the business type indicating the inquiry destination of the user, and the identification information of the dictionary used for voice recognition;
A dictionary storage unit in which a dictionary for voice recognition for each operator and a dictionary for voice recognition for each business type are stored;
A voice storage unit that stores the voice of the operator corresponding to the telephone;
A call information storage unit that stores call information in which identification information of an operator corresponding to a telephone call is associated with identification information of a business type handled by the operator;
A dictionary selection unit that selects a dictionary to be used for speech recognition with reference to the dictionary management information storage unit based on the call information;
A voice recognition device comprising: a voice recognition unit that converts the operator's voice read from the voice storage unit into text data with reference to the dictionary in the dictionary storage unit selected by the dictionary selection unit. The dictionary selection unit
Select dictionary identification information to be used for speech recognition with reference to the dictionary management information storage unit based on the call information,
The voice recognition unit
The voice recognition according to claim 1, wherein the voice of the operator recorded by the call recording device is converted into text data with reference to the dictionary of the dictionary storage unit that matches the identification information of the dictionary selected by the dictionary selection unit. apparatus.   2. The speech recognition apparatus according to claim 1, wherein the dictionary for speech recognition for each operator is a dictionary in which text data and speech data are registered in association with each operator's pronunciation and speaking style.   The speech recognition apparatus according to claim 1, wherein the speech recognition dictionary for each business type is a dictionary in which text data and speech data of technical terms for each business type are linked and registered. In addition to a dictionary for voice recognition for each operator and a dictionary for voice recognition for each business type, the dictionary storage unit includes a general-purpose dictionary,
The speech recognition apparatus according to claim 1, wherein the dictionary selection unit selects the general-purpose dictionary as a dictionary used for speech recognition when there is no dictionary that matches the call information in the dictionary storage unit. A dictionary management information storage unit that stores the identification information of the operator corresponding to the user, the identification information of the business type indicating the inquiry destination of the user, and the identification information of the dictionary used for voice recognition;
A dictionary storage unit in which a dictionary for voice recognition for each operator and a dictionary for voice recognition for each business type are stored;
A call recording device for recording the voice of an operator who corresponds to the user by telephone;
A call information generation unit that generates call information in which identification information of an operator corresponding to a telephone call is associated with identification information of a business type handled by the operator;
A dictionary selection unit that selects a dictionary to be used for speech recognition with reference to the dictionary management information storage unit based on the call information;
A voice recognition system comprising: a voice recognition unit that converts a voice of an operator recorded by the call recording device into text data with reference to the dictionary of the dictionary storage unit selected by the dictionary selection unit. An identifier of an operator corresponding to the user, a business type identification information indicating a user's inquiry destination, and a dictionary used for voice recognition are stored in association with each other,
Stores a dictionary for voice recognition for each operator and a dictionary for voice recognition for each business type,
Store or record the voice of the operator who is on the phone,
Stores call information that associates the identification information of the operator corresponding to the user with the identification information of the business type that the operator is in charge of,
A speech recognition method for selecting a dictionary to be used for speech recognition based on the call information from a plurality of stored dictionaries, and referring to the selected dictionary to convert stored or recorded operator speech into text data .

Images