KR20100058687A - Apparatus and method for processing speech recognition for large vocabulary speech recognition - Google Patents

Abstract

PURPOSE: A speech recognition device for the bulk voice recognition and a method thereof are provided to perform the bulk voice recognition in a limited computing environment. CONSTITUTION: A voice recognition support unit(120) receives voice recognition target words from a user according to the scenario of a voice recognition service through a server interface unit(140). If the voice recognition target word is inputted, a recognition target address book management unit(110) creates an address book integrated by a personal address book and a business address book. A voice recognition unit(130) recognizes the voice recognition target word as the voice recognition name of the integration address book.

Description

Speech Recognition Processing Apparatus and Method for Massive Speech Recognition {APPARATUS AND METHOD FOR PROCESSING SPEECH RECOGNITION FOR LARGE VOCABULARY SPEECH RECOGNITION}

The present invention relates to a voice recognition processing apparatus and a method for large-scale voice recognition, and more particularly, to generate a user-specific integrated address book using a personal address book and a corporate address book for each user, the user's voice recognition target words in the generated integrated address book By recognizing the voice recognition name corresponding to, it is possible to provide a stable voice recognition service to the user by performing a large amount of voice recognition accurately and quickly even in a restricted computing environment (for example, a 1,000 word speech recognition engine). The present invention relates to a speech recognition processing apparatus and a method for mass speech recognition.

Voice recognition technology is applied to a voice recognition service that provides a voice recognition function as a user interface, and is still developing. However, voice recognition technology has difficulty in providing a service that satisfies the level desired by the user in the market. A description will be given using "brand call service" as an example. In the case of "brand call service", service activation was limited due to voice recognition performance. At the time, there was a demand from the user that the voice recognition rate should be 100%. However, it is now recognized as a convenient interface rather than having to be 100%. In addition, the demand for speech recognition has become more specific. There is an urgent need for a large capacity speech recognition method that reflects these user requirements. This problem must be solved in the voice recognition technology. If this problem is not solved, it is difficult to efficiently provide a large capacity voice recognition service to a user who wants a large capacity voice recognition.

On the other hand, even if there is a voice recognition engine that provides a large capacity voice recognition service, the voice recognition function is limited by the system computing ability. The voice recognition function is a CPU-bound job (CPU Bound Job), mainly the floating point (FP) operation. When several channels simultaneously occupy the central processing unit and require voice recognition, the CPU share in the voice recognition server reaches 100%. Speech recognition algorithms use a lot of CPU resources.

Therefore, even if a voice recognition system adopts a high performance speech recognition engine, it is difficult to process a large volume of speech recognition due to the limitation of computing power. In particular, a service over IP (SoIP) system that utilizes a broadband convergence network (BcN) as a base network needs a voice recognition system that overcomes some limitations of computing power. In addition, Internet telephony systems require large-scale sound ceremony.

There is a problem that it is difficult to provide a large capacity voice recognition service in the conventional voice recognition system having a limited computing capability. For example, voice recognition servers in the IP Multimedia Subsystem (IMS) and Broadband Integrated Network (BcN) media server areas, introduced several years ago and currently without system upgrades, are specific. This can be an example. Such voice recognition server will actively reflect on Internet telephony technology using broadband integrated network (BcN).

The enumerated intelligent network and broadband integrated network (BcN) equipment has been introduced in accordance with the service situation over the years, so that the performance difference between the computing capabilities according to the introduction time is significantly increased. For example, the recently introduced system's CPU is a quad core with four cores, but the initial product is a single core of 800 MHz. Therefore, due to the constraints of the system purchase specification, the existing service plans to develop the service at 1,000 words if the system purchase standard is 1,000 words. However, the latest equipment can handle more than 1,000 words, giving you more freedom in service planning.

By way of example of the "brand call service" described above, a brief description of the "brand call service" scenario is as follows.

First, the user accesses the brand call service access number.

And the brand name of the company the customer wants to talk through the user terminal. For example, the user may say "Company A".

Then, the voice recognition service system performs voice recognition.

Then, if the customer wants a voice recognition result, the voice recognition service system performs a call to the company.

In the voice recognition service scenario, the voice recognition target is a brand name of a company. The user of the voice recognition service may think that not only 1,000 words are recognized when using the service. In other words, he understands that the habit of using a service is all he wants. Therefore, the user's position using the service seems to lack quality when looking at the satisfaction from the voice recognition perspective of the "brand call service". In this case, the "brand call service" has a problem that does not provide the convenience of use. Thus, the "brand call service" is turned away from users. That is, since the conventional technology is speech recognition on the scale of 1,000 words, only the brand name corresponding to 1,000 words is selected and the voice recognition for the selected brand name is provided to the majority of users. Therefore, the conventional speech recognition technology has a problem in securing an appropriate speech recognition rate. Therefore, the conventional voice recognition technology has a problem in that the user receives an external appearance because the quality of the service is lowered.

Therefore, the prior art as described above has a problem in that a user cannot efficiently provide a large capacity speech recognition service to a user in a limited computing environment (for example, a 1,000 word speech recognition engine). It is a subject of the present invention.

Therefore, the present invention generates a consolidated address book using a personal address book and a corporate address book for each user, and recognizes a voice recognition name corresponding to a user's voice recognition word in the generated integrated address book, thereby restricting computing environment (for example, It is an object of the present invention to provide a speech recognition processing device and a method for mass speech recognition, which can provide a stable speech recognition service to a user by performing a large volume of speech recognition accurately and quickly even in a 1,000 word class speech recognition engine). have.

The objects of the present invention are not limited to the above-mentioned objects, and other objects and advantages of the present invention which are not mentioned can be understood by the following description, and will be more clearly understood by the embodiments of the present invention. Also, it will be readily appreciated that the objects and advantages of the present invention may be realized by the means and combinations thereof indicated in the claims.

The present invention is to solve the above problems, characterized in that for generating a user-specific integrated address book and a corporate address book for each user, and to recognize the voice recognition name corresponding to the user's voice recognition target words in the generated integrated address book .

More specifically, the present invention provides a speech recognition processing apparatus for a large capacity speech recognition, comprising: server interface means for providing an interface for a speech recognition service with an application server; Voice recognition support means for receiving a voice recognition target word from a user according to a scenario of a voice recognition service through the interface to support a voice recognition service; Recognition target address book management means for separately managing the personal address book and the corporate address book, for generating an integrated address book for each user by using the managed personal address book and the corporate address book when the voice recognition target word is input from the user by voice; And voice recognition means for recognizing the input voice recognition target word as a voice recognition name in the generated integrated address book.

On the other hand, the present invention, in the speech recognition processing method for large-capacity speech recognition, word input step of receiving a voice recognition target word from the user according to the scenario of the speech recognition service through the interface with the application server; Generating an integrated address book for each user by using a personal address book and a corporate address book distinguished from each other when a voice recognition target word is input by voice from the user; And a voice recognition step of recognizing the input voice recognition target word as a voice recognition name in the generated integrated address book.

In addition, the method of the present invention further includes a customer preference extraction step of extracting customer preferences by analyzing the usage history of the recognized voice recognition name for each user.

The present invention as described above, by creating an integrated address book using a personal address book and a corporate address book for each user, by recognizing a voice recognition name corresponding to the user's voice recognition words in the generated integrated address book, limited computing environment (eg For example, the 1,000 words speech recognition engine) has the effect of providing a stable speech recognition service to the user by performing a large volume of speech recognition accurately and quickly.

That is, the present invention, by providing a large volume of voice recognition in the service using the Internet phone terminal of the BcN-based network through the system of the existing standard, through the cost reduction and guarantee of the voice recognition rate according to the system construction, large capacity in the service development There is an effect that can enable the provision of speech recognition. In addition, the present invention provides accurate voice recognition performance and stability when developing a service for applying voice recognition, thereby providing convenience for providing a stable service to a user, thereby contributing to the activation of a business.

BRIEF DESCRIPTION OF THE DRAWINGS The above and other objects, features and advantages of the present invention will become more apparent from the following detailed description of the present invention when taken in conjunction with the accompanying drawings, It can be easily carried out. In the following description, well-known functions or constructions are not described in detail since they would obscure the invention in unnecessary detail. Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram of a broadband integrated network system to which the speech recognition processing apparatus according to the present invention is applied.

As shown in FIG. 1, the broadband integrated network system includes a voice recognition processing apparatus 100 and an application server 10. Here, the voice recognition processing apparatus 100 includes a recognition target address book management unit 110, a voice recognition support unit 120, a voice recognition unit 130, and a server interface unit 140. In addition, the subscriber terminal 101 includes an enterprise customer terminal 1011 and an individual customer terminal 1012.

The speech recognition processing apparatus 100 is applied to a broadband integrated network system using a server-based speech recognition service as shown in FIG. 1. That is, an example applied in the environment of a broadband integrated network is shown in FIG. In addition, the voice recognition processing apparatus 100 may be implemented as shown in FIG. 1 in the case of an intelligent network.

Here, in order to process the large-capacity speech recognition in the speech recognition processing apparatus 100, it is basically necessary to improve the performance of the speech recognition engine. The speech recognition unit 130 of the speech recognition processing apparatus 100 includes an automatic speech recognition (ASR) server, and provides a speech recognition function using a speech recognition engine. Even if the specification of the voice recognition engine is provided with a low voice recognition server, the voice recognition processing apparatus 100 according to the present invention can be usefully applied when a large volume of voice recognition is desired.

On the other hand, when a user requests a voice recognition service using a voice recognition service number through a user terminal 102 (for example, a broadband integrated network (BcN) terminal), the server interface unit of the voice recognition processing apparatus 100 ( 140 transmits the user's request to the application server (10).

The application server 10 checks the static grammar of the voice recognition service number received from the server interface 140 of the voice recognition processing apparatus 100 and the location of the VXML url. The application server 10 transmits the verification result to the voice recognition processing apparatus 100 through the server interface unit 140.

Subsequently, the speech recognition processing apparatus 100 provides a speech recognition service based on a voice eXtensible Markup Language (VXML) to the user terminal 102. The voice input by the user through the user terminal 102 is recognized by the voice recognition processing apparatus 100, and the recognition result is transmitted to the application server 10 through the server interface unit 140.

Thereafter, the application server 10 provides the application service to the user through the user terminal 102 by using the recognition result.

Hereinafter, a process of registering a voice recognition target address book from a subscriber who wants to provide a voice recognition service through the voice recognition processing apparatus 100 according to the present invention, that is, the owners of the voice recognition target address book will be described.

For speech recognition, the speech recognition unit 130 must perform a speech recognition initialization process using the speech recognition target word in the speech recognition engine provided. That is, the speech recognition unit 130 may recognize speech only words known to the speech recognition engine. For example, suppose the target word for speech recognition is a company name and a large amount (for example, 10,000 words). If this is assumed, the conventional voice recognition server selects only 1,000 words which are most likely to be used by the user, and initializes the voice recognition, and provides the voice recognition result to the user through the user's telephone connection.

However, assuming that the voice recognition unit 130 according to the present invention recognizes the corporate name, the corporate customer can use the corporate customer terminal 1011 to input a voice recognition target address book (hereinafter, referred to as a corporate address book) for the corporate name. (For example, the service web) is registered in the recognition target address book management unit 110. That is, the corporate customer who wants to recognize the voice must register the company name on the web through the corporate customer terminal 1011 to register the desired name in the corresponding enterprise. This allows the user to use the correct company name for future services. In addition, this allows the voice recognition unit 130 to correctly recognize the recognition rate of the company name.

On the other hand, the recognition target address book management unit 110 receives a personalized service list from the personal customer terminal 1012 of all individual customers who want to use the voice recognition service. Here, the personalized service list is composed of a plurality of company names that the individual customer wants to receive the voice recognition service. When the voice recognition target address book (hereinafter, referred to as a 'personal address book') desired by the individual is registered in the recognition target address book management unit 110, the application server 10 may provide a personalized service to the individual by using the registered personal address book. do. In addition, the recognition target address book management unit 110 manages an integrated address book that can increase the voice recognition rate through a personalized service. In order to prevent the unspecified number from being used without any limitation as in the conventional "brand call service", the recognition target address book management unit 110 registers a personal address book from the personal customer terminal 1012 of the individual customer who wants to receive a voice recognition service. do. In this case, if the service is provided by a non-specified number of individual customers who want to use personally, voice recognition performance is degraded as in the conventional voice recognition technology.

Meanwhile, a description will be given of a voice recognition initialization process performed by the voice recognition unit 130.

The conventional speech recognition server initializes a target word for speech recognition. The initialization of the target word performed in the conventional voice recognition server may be divided into fixed initialization and dynamic initialization according to the type of service. Fixed initialization means that the voice recognition server once initialized does not initialize again. In contrast, dynamic initialization means that the voice recognition server dynamically initializes when necessary.

On the other hand, the speech recognition unit 130 performs the mixed speech recognition initialization by mixing the fixed type and the dynamic initialization described above. That is, the voice recognition unit 130 integrates the corporate address book which is not fixedly changed and the personal address book which is frequently used by the individual customer as the user and stores the integrated address book. The voice recognition unit 130 performs a fixed initialization of the integrated address book, that is, the corporate address book, and performs a dynamic initialization of the personal address book.

On the other hand, even if the speech recognition unit 130 is provided with a low speech recognition engine to improve the speech recognition rate when a large volume of speech recognition request. Therefore, the voice recognition support unit 120 collects the customer preferences for the user history using the voice recognition through the user terminal 102 through the recognition result in the voice recognition unit 130. In addition, the recognition target address book management unit 110 stores the collected customer preferences in the corresponding address book and reflects them in each address book.

2 is a block diagram of an embodiment of a speech recognition processing apparatus for mass speech recognition according to the present invention.

As shown in FIG. 2, the voice recognition processing apparatus 100 includes a recognition target address book management unit 110, a voice recognition support unit 120, a voice recognition unit 130, and a server interface unit 140. Here, the recognition target address book management unit 110 includes an address book management unit 111, a personal address book storage unit 112, a corporate address book storage unit 113, a customer preference storage unit 114, and an integrated address book management unit 115. Here, the customer is divided into an individual customer (a customer who uses a voice recognition service in the future) and a corporate customer (a company that wants voice recognition) for the registration process of the voice recognition object name and the voice recognition initialization process. In addition, the customer preference is stored in the customer preference storage unit 114 according to a commonly used address book and customer usage history. The personal address book will be updated according to the customer's preferred name.

Hereinafter, each component of the speech recognition processing apparatus 100 will be described in detail.

First, a corporate customer registers a corporate name on the web by inputting a company name, which is a voice recognition target word, and a corresponding telephone number, through the corporate customer terminal 1011. In addition, the address book manager 111 registers a company name necessary for a voice recognition service inputted by the corporate customer through the corporate customer terminal 1011, that is, a company name which is a voice recognition target word through the web. Here, the address book manager 111 stores the registered company address book in the company address book storage 113.

In addition, the individual customer enters the company name, which is a voice recognition target word of his or her preference, through the personal customer terminal 1012 and registers on the web. In addition, the address book manager 111 classifies the company name registered by the personal customer terminal 1012 for each individual and stores the company name in the personal address book storage 112. Individual customers will have a personal address book upon signing up. This personal address book can be used for personal voice recognition.

In other words, the corporate customer performs service registration on the web through the corporate customer terminal 1011. The corporate customer registers the service by inputting the recognition name and the incoming telephone number desired by the corporate customer. And individual customers who want to receive voice recognition service also register service through service web.

The address book manager 111 stores the recognition name and the received telephone number registered from the corporate customer terminal 1011 in the corporate address book storage 113 to make a database. In addition, the address book manager 111 stores the recognition name and the received telephone number transmitted from the personal customer terminal 1012 in the personal address book storage 112 to make a database.

When the user requests the voice recognition service through the user terminal 102, the address book manager 111 constructs an integrated address book by integrating the company name corresponding to the voice recognition target address book (personal address book and corporate address book). For example, the address book manager 111 may manage 500 recognition names in the rank of the most preferred company name and 500 recognition names which can be modified in the personal address book.

Meanwhile, at the request of the individual customer, the address book manager 111 may update the personal address book stored in the personal address book storage 112. However, about 500 company names fixedly stored in the corporate address book storage 113 cannot be modified by individual customers. The 500 company names are fixed data in voice recognition initialization. The remaining 500 company names can be directly managed by individual customers through personal address books. That is, when an individual customer accesses the network to use the service, the address book manager 111 integrates the voice recognition target address book of the speech recognition engine by adding 500 fixed words and 500 words that can be modified by the individual through user authentication. To build an integrated address book.

As described above, the voice recognition processing apparatus 100 intends to provide a large capacity voice recognition service to a registered individual customer through a service web or the like. That is, the individual customer performs user registration on the service web. In this case, a large-capacity voice recognition service may be provided using the personal customer terminal 1012 registered on the service web. In this case, the individual customer may be provided with a voice recognition service at a high voice recognition rate even in the limited computing environment of the voice recognition unit 130.

That is, the voice recognition processing apparatus 100 is to provide a large capacity voice recognition service that provides a high voice recognition rate even through a low voice recognition engine in a broadband integrated network (BcN) -based Internet telephone (SoIP). Therefore, in providing an Internet telephony (SoIP) based voice recognition service in a broadband integrated network, the voice recognition processing device 100 is to be applied to a system for providing a large capacity voice recognition.

On the other hand, in order to provide a voice recognition service to the individual customer, the address book management unit 111 builds a usage record analyzed by the individual customer history in the customer preference storage unit 114 as a database. Accordingly, the voice recognition support unit 120 extracts customer preferences from a pre-built database. In addition, when the voice recognition support unit 120 delivers the extracted customer preferences to the address book manager 111, the address book manager 111 may reflect the extracted customer preferences to the personal address book at random intervals. The address book manager 111 may update the integrated address book stored in the integrated address book manager 115 previously constructed using the updated personal address book. Here, the personal address book which can be modified for each individual customer other than the predetermined fixed corporate address book is updated.

Then, when the user requests the voice recognition service 100 through the user terminal 102, the voice recognition support unit 120 checks the user to the address book management unit 111 to the integrated address book for the user request. Then, the address book manager 111 builds an integrated address book for the user in the integrated address book by using the personal address book of the user who requested voice recognition and a predetermined corporate address book.

Subsequently, the voice recognition support unit 120 receives the integrated address book for the user from the recognition target address book management unit 110 and transmits the received address book to the voice recognition unit 130. The voice recognition unit 130 performs a voice recognition initialization process using the integrated address book for the received user. At this time, the voice recognition support unit 120 checks the number of the registered calling terminal when the caller subscribed to the service, and delivers the integrated address book for the user of the identified calling terminal to the voice recognition unit 130 to initialize the voice recognition To perform Accordingly, the voice recognition unit 130 may provide the user with a customized service by initializing the voice recognition target address book using the integrated address book according to the personal address book registered in advance for each individual. This is to support a large capacity voice recognition service without changing the voice recognition performance even though the hardware specification of the conventional voice recognition engine is low. In addition, it is possible to improve the recognition performance by providing a large capacity voice recognition service in a given environment. Therefore, it can greatly contribute to the activation of broadband integrated network SoIP technology.

On the other hand, the performance of the speech recognition unit 130 is greatly influenced by hardware (speech recognition engine) specifications. This is because the voice recognition algorithm is processed based on more resources of the CPU than other functions. Furthermore, when the target word for speech recognition increases in large capacity, the speech recognition time may increase rapidly. Therefore, the voice recognition processing apparatus 100 having the capability of 1,000 words of voice recognition words has more influence on voice recognition time than voice recognition rate as the words to be recognized are increased.

Therefore, the voice recognition processing apparatus 100 according to the present invention intends to provide a user with a large voice recognition effect on a limited hardware specification. To this end, the address book manager 111 constructs an integrated address book by using a corporate address book registered by a corporate customer, a personal address book registered by a personal customer, and customer preferences reflecting user usage details. Then, the speech recognition processing apparatus 100 provides the user with the effect of the large capacity speech recognition service while ensuring the speech recognition rate. The speech recognition processing apparatus 100 may provide a speech recognition service through a recognition name which is generally not new to general users but generally used. In addition, the voice recognition processing apparatus 100 may provide a voice recognition service to which the voice recognition rate is guaranteed by providing a voice recognition service through reflection according to the user's customer preference.

On the other hand, in the voice recognition service through the company name, there is a difference between the name of the company that is actually used administratively and the name of the name commonly used by real users. For the sake of simplicity, the name of the company has been reduced in detail so that the name of the company represents the company and implies meaning in the recognition name. But the average user tends to speak as simply as possible. For example, "Samsung Fire & Marine Insurance Co., Ltd." is the official name, but users generally refer to the brand "Samsung Fire" or "Anika".

Therefore, in the case of speech recognition processing of the company name, the recognition name that general users talk about the company may be collected in advance. In addition, the address book manager 111 may enable the user to directly input the collected recognition names, thereby increasing the voice recognition rate and increasing customer satisfaction with the corresponding service.

On the other hand, when the voice recognition unit 130 recognizes the name of the company, if the target of the unspecified number of voice recognition performance is low, voice recognition service becomes difficult. It is difficult to provide the same voice recognition service to an unspecified number because there are too many words to be recognized and new company names increase.

Therefore, the voice recognition processing device 100 is to provide a customized to the customer using the voice recognition service. The address book manager 111 registers a service on the web from an individual customer who wants to receive a company name by voice recognition. In addition, corporate customers enter the company name they want.

Meanwhile, the voice recognition unit 130 may not provide a voice recognition function for all the selected company names, and voice recognition may be performed for the company name initialized by the voice recognition support unit 120 by the voice recognition support unit 120. The address book manager 111 selects 1,000 words by checking the frequency of the recognition target name analyzed in advance. For example, since the hardware specification of the voice recognition engine is 1,000 words for the voice recognition service, the address book manager 111 builds an integrated address book by dividing the fixed company name and the dynamic company name among the selected words. The address book manager 111 stores the five hundred words which are 50% of the company names stored in the company address book storage 113 in the integrated address book manager 115 so that the general user cannot modify them. The remaining 50% of the five hundred words are filled with the remaining 500 words of 1,000 words, allowing the general user to modify them. At this time, in order to be consistent with the company name registered by the enterprise, the address book of the recognition target word is made of the following names.

Specifically, the address book manager 111 prevents the user from modifying the company name registered by the first corporate customer. That is, the telephone number for each company name directly entered by the corporate customer cannot be modified by the user. On the other hand, the address book manager 111 enables the user to modify the company name desired for the voice recognition name that can be modified. In addition to the fixed five hundred words out of 1,000 words, the user can modify the five hundred words user can modify. In addition, the user registers on the web with the user's outgoing call.

Meanwhile, when the general user who registers the voice recognition service uses the voice recognition service, the address book manager 111 receives the user usage history from the voice recognition supporter 120 and stores the user usage history in the customer preference storage 114. In the case of a user who subscribes to a voice recognition service, in order to grasp voice recognition performance and usage history, the customer preference storage unit 114 builds a database and stores customer preferences. Here, the customer preference storage unit 114 stores the calling terminal number (CID), user voice information, and voice recognition result as customer preferences. The calling terminal number (CID) is stored in the customer preference storage unit 114 for user identification.

The speech recognition processing apparatus 100 performs a process of extracting customer preferences at random intervals from a database constructed as an integrated address book and reflecting them back to the speech recognition service.

In detail, the address book manager 111 ranks each speech name by synthesizing the speech names of all users who request the speech recognition service delivered through the speech recognition support 120.

And the address book management unit 111 has a ranked ranking and reflects newly to the corporate address book consisting of a fixed 500 address book for the voice recognition name from the first place to the 500th place. The address book manager 111 reflects the remaining 500 names in a personal address book that can be modified by a customer. Here, the address book manager 111 may reflect only the newly entered entry in the personal address book in comparison with the voice recognition target address book which the user may modify before. In addition, the address book manager 111 notifies the user of the newly entered address book. Then, the user may modify the previous personal address book based on what is known through the address book manager 111. This is to reflect customer preferences in an individual's address book.

Thereafter, the voice recognition target address book, which was used for the initial voice recognition service start, is changed over time through the usage history of actual users as time passes. Thus, over time, corporate names (recognition names) will be refined according to personal preferences and usage details. Therefore, since a large amount of voice recognition requirements are provided to the individual customization, the address book manager 111 manages a personal address book, corporate address book and integrated address book in order to provide a voice recognition performance of the level desired by the user.

When the voice recognition service is provided to the user, the voice recognition unit 130 uses the voice recognition names (company name) updated by the recognition target address book management unit 110. Therefore, the address book manager 111 sets a terminal number registered at the time of service subscription for each user as one identification number and stores the personal address book for each user in the personal address book storage 112. The personal address book can be modified by the user through the personal customer terminal 1012 on the web, and is reflected in the personal address book storage 112 in real time when the modification is made.

Figure 3 is a flow diagram of an embodiment of a speech recognition processing method for large-capacity speech recognition according to the present invention.

First, the voice recognition support unit 120 receives a voice recognition target word from the user terminal 102 according to the scenario of the voice recognition service through the interface with the application server 10 (302).

When the voice recognition target word is input from the user terminal 102 by voice, the recognition target address book management unit 110 generates an integrated address book for each user by using the personal address book and the corporate address book separated from each other (304).

Subsequently, the voice recognition unit 130 recognizes the input voice recognition target word as a voice recognition name in the integrated address book generated by the recognition target address book management unit 110 (306).

Thereafter, the voice recognition support unit 120 extracts the customer preference by analyzing the usage history of the voice recognition name recognized by the voice recognition unit 130 for each user (308).

The recognition target address book management unit 110 includes the user's calling terminal number, voice data and the recognized voice recognition name in the customer preference extracted by the voice recognition support unit 120, and uses the extracted customer preference to create a personal address book. Update 310.

Meanwhile, the recognition target address book management unit 110 checks the frequency of the recognized voice recognition names, selects a predetermined number of voice recognition names, and generates an integrated address book including the selected voice recognition names. In addition, the recognition target address book management unit 110 generates the integrated address book by dividing the selected voice recognition name into a name that can be modified at a preset rate and a name that cannot be modified.

On the other hand, the method of the present invention as described above can be written in a computer program. And the code and code segments constituting the program can be easily inferred by a computer programmer in the art. In addition, the written program is stored in a computer-readable recording medium (information storage medium), and read and executed by a computer to implement the method of the present invention. The recording medium may include any type of computer readable recording medium.

The present invention described above is capable of various substitutions, modifications, and changes without departing from the technical spirit of the present invention for those skilled in the art to which the present invention pertains. It is not limited by the drawings.

1 is a block diagram of a broadband integrated network system to which a speech recognition processing device according to the present invention is applied;

2 is a block diagram of an embodiment of a speech recognition processing apparatus for mass speech recognition according to the present invention;

3 is a flowchart illustrating an embodiment of a speech recognition processing method for mass speech recognition according to the present invention.

* Explanation of symbols for the main parts of the drawings

10: application server 100: speech recognition processing unit

110: address book management unit 111: address book management unit

112: personal address book storage 113: corporate address book storage

114: customer preference storage 115: integrated address book management unit

120: speech recognition support unit 130: speech recognition unit

140: server interface unit

Claims (13)

In the speech recognition processing apparatus for a large capacity speech recognition, Server interface means for providing an interface for voice recognition service with an application server; Voice recognition support means for receiving a voice recognition target word from a user according to a scenario of a voice recognition service through the interface to support a voice recognition service; Recognition target address book management means for separately managing the personal address book and the corporate address book, for generating an integrated address book for each user by using the managed personal address book and the corporate address book when the voice recognition target word is input from the user by voice; And Voice recognition means for recognizing the input voice recognition target word as a voice recognition name in the generated integrated address book Speech recognition processing device comprising a. The method of claim 1, The voice recognition support means, And a function of extracting customer preferences by analyzing usage history of the recognized voice recognition name for each user. The method of claim 2, The voice recognition support means, And a voice calling terminal number, voice data, and the recognized voice recognition name of the user in the extracted customer preferences. The method of claim 3, wherein The recognition target address book management means, Further comprising customer preference storage means for storing the extracted customer preferences for each user, Voice recognition processing device for updating the managed personal address book using the stored customer preferences. The method according to any one of claims 1 to 4, The recognition target address book management means, Address book management means for registering a personal address book and a corporate address book from a personal customer and a corporate customer, respectively, and generating and managing an integrated address book by integrating the registered personal address book and the registered corporate address book; Personal address book storage means for storing the personal address book registered by the personal customer for each customer; Corporate address book storage means for storing a corporate address book registered by the corporate customer; And Integrated address book storage means for storing the generated integrated address book Speech recognition processing device comprising a. The method of claim 5, The address book management means, And a predetermined number of speech recognition names are selected by checking the frequency of the recognized speech recognition names and generating an integrated address book including the selected speech recognition names. The method of claim 6, The address book management means, And generating the integrated address book by dividing the selected voice recognition name into a name that can be modified and a name that cannot be modified at a preset ratio. The method of claim 5, The address book management means, And analyzing the recognized voice recognition name to obtain a rank for the voice recognition name, and newly reflecting the obtained rank in the voice recognition name of the stored personal address book. In the speech recognition processing method for large-capacity speech recognition, A word input step of receiving a voice recognition target word from a user according to a scenario of a voice recognition service through an interface with an application server; Generating an integrated address book for each user by using a personal address book and a corporate address book distinguished from each other when a voice recognition target word is input by voice from the user; And A voice recognition step of recognizing the input voice recognition target word as a voice recognition name in the generated integrated address book Speech recognition processing method comprising a. The method of claim 9, Customer preference extraction step of extracting customer preferences by analyzing the use history of the recognized voice recognition name for each user Speech recognition processing method further comprising. The method of claim 10, The customer preference extraction step, And including the user's calling terminal number, voice data, and the recognized voice recognition name in the extracted customer preferences, and updating the managed personal address book using the extracted customer preferences. The method according to any one of claims 9 to 11, The integrated address book generation step, And a predetermined number of speech recognition names are selected by checking the frequency of the recognized speech recognition names, and generating an integrated address book including the selected speech recognition names. 13. The method of claim 12, The integrated address book generation step, And generating a unified address book by dividing the selected voice recognition name into a name that can be modified and a name that cannot be modified at a preset ratio.

Images