KR20130127901A - Apparatus and method for speech recognition - Google Patents

Abstract

The present invention relates to a device for speech recognition and automatic interpretation operating on a PC or mobile device. The speech recognition device according to the present invention is a screen for selecting a domain as a unit for a speech recognition region of a predetermined classification for speech recognition. Display unit for displaying to the user; A user input unit for receiving a selection of a domain from the user; And a communication unit transmitting the selection information of the user for the domain. According to the present invention, a voice recognition device is provided to the user through an intuitive and simple user interface, so that the user can easily select / modify the designated domain of the voice recognition system, and the accuracy of voice recognition and automatic interpretation through the designated voice recognition system. It can improve performance.

Description

 [0001] Apparatus and method for speech recognition [0002]

The present invention relates to a device or a speech recognition method equipped with a speech recognition and automatic interpretation function, and more particularly, to a method of selecting a domain of a database for speech recognition.

Conventional speech recognition or automatic interpretation systems are inefficient to train all vocabulary and expressions in various fields, so they are usually trained in only one domain, that is, domain. Conventional speech recognition or automatic interpretation applications in most cases can not modify the domain specified by default. In addition, even when the user can directly select the domain, there is a problem that the user is inconvenient to use and the selection is very simple. Therefore, there is a problem that the adaptability to the speech recognition environment is lowered, and the accuracy of speech recognition and automatic interpretation is lowered.

The present invention aims to increase the accuracy of speech recognition and automatic interpretation by providing a user interface that allows users to easily select a database, that is, a domain referred to in speech recognition or automatic interpretation, to easily select a domain according to a situation. It is done.

According to an aspect of the present invention, there is provided a speech recognition apparatus, including: a display unit configured to display a screen for selecting a domain to a user as a unit for a speech recognition region having a predetermined classification for speech recognition; A user input unit for receiving a selection of a domain from the user; And a communication unit transmitting the selection information of the user for the domain.

According to the present invention, it is possible to provide a user with an intuitive and simple method for selecting a domain, thereby improving the accuracy and performance of voice recognition and automatic interpretation.

1 is a network diagram for providing a voice recognition service according to an embodiment of the present invention.
2 shows a configuration of a user terminal according to an embodiment of the present invention.
3 exemplarily shows a structure and a relationship of a designated domain according to an embodiment of the present invention.
4 to 11 illustrate screens displayed on a display unit of a user terminal according to an exemplary embodiment of the present invention.
12 is a flowchart illustrating a voice recognition domain designation method according to an embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings.

1 is a block diagram illustrating a network for providing a voice recognition method according to an embodiment of the present invention.

The user terminal 10 receives voice and domain selection information from the user and delivers the voice and domain selection information to the voice recognition server 20. The user terminal 10 may be any computing device equipped with a communication function such as a PC, a notebook computer, a smartphone, and the like, which allows a user to input voice or text.

The voice recognition server 20 performs voice recognition by referring to data corresponding to the domain selected by the user among the voice recognition reference data stored in the DB 30 through the received voice and information on the selected domain. Then, the performed voice recognition result is transmitted to the user terminal 10.

In the DB 30, various data necessary for the voice recognition operation of the voice recognition server 20 are stored, and data to be referred to during the voice recognition operation, such as a corpus and a language dictionary, are stored for each domain.

Hereinafter, the user terminal 10 will be described in more detail with reference to FIG. 2.

As illustrated in FIG. 2, the user terminal 10 according to the present exemplary embodiment may include a display unit 100, a user input unit 200, and a communication unit 300.

The display unit 100 displays information necessary for speech recognition and may display menus for designating a domain to be referred for speech recognition to a user. In the present embodiment, the speech recognition server 20 is a system that recognizes meaning by receiving a speech signal and performs speech recognition based on a designated domain or a general domain designated by a user.

A general domain is a database that is referred to to support speech recognition for commonly used languages rather than a specific domain, and a designated domain is automatically applied to a specific situation or to a user in order to support more accurate speech recognition than the general domain described above. Selected by the database. For example, if the input voice is related to travel, voice recognition may be performed using the 'travel' domain as a designated domain, and may generate better voice recognition results than when the general domain is selected.

The concept of the speech recognition domain will be described in more detail with reference to FIG. 3. Also, in the present embodiment, the speech recognition domain may be referred to as a unit for classifying a region of speech recognition, that is, a database referred to during a speech recognition process.

Referring to FIG. 3, as described above, the voice recognition server 20 may operate with respect to the general domain 31 by default or by user selection. Each designated domain may have a first subdomain 32 and second subdomains 33 having the first subdomain 32 as a parent domain. Further, although not shown, it may include third and fourth subdomains having the second subdomain as a parent domain.

These second subdomains may replace some characteristic (word or expression) of the parent domain or may have a characteristic that is not in the parent domain. In addition, the respective domains may overlap each other. For example, two subdomains, such as a touring domain, may partially overlap with another domain, the business domain, and a portion of the restaurant domain, a subdomain of the travel domain, may overlap with the business domain. Can be.

Hereinafter, a configuration of a screen for selecting a domain displayed on the display unit 100 according to the present embodiment will be described with reference to the drawings.

In the present embodiment, the domain display unit 100 displays a domain that the user can select or deselect. Referring to FIG. 4, the display 100 may display domains in a hierarchical or tree structure. As shown in FIG. 4, each domain may be indicated by a label which is the name of a designated domain of the domain.

The general domain is represented by a label named 'General', and may include a travel domain 47, a business domain, a meeting domain 45, and a medical domain 46 as four subdomains of the general domain. have.

The travel-related domain is represented by a domain 47 having a label named 'Touring', and may further include three sub-domains, a restaurant domain, an airport domain, and a car rent domain. . Furthermore, the restaurant domain is represented as a domain having a label named 'Restaurant' and may include additional subdomains according to the type of restaurant. For example, in the case of Figure 4, the Korean restaurant domain is represented by the label 'Korean Food', the Chinese restaurant domain 'Chinese Food'.

The Korean restaurant domain may include language data for Korean food and restaurant names to support speech recognition.

In addition, the conference domain may be represented by a label 45 named 'Conference', and may include a computer engineering domain and a mechanical engineering domain as subdomains. In the case of a computer engineering domain, it may be represented by a label of 'Computer Science', and the mechanical engineering domain may be represented by a label of 'Mechanical Engineering'. In the case of conference domains, the use of professional vocabulary is higher than in other areas. Therefore, in case of providing a designated voice recognition service by subdividing the voice recognition area according to the related field of conference, the accuracy of recognition and further interpretation accuracy can be improved. .

Hereinafter, the user input unit 200 that receives a selection of a domain for selecting a domain from a user through the domain display unit 100 will be described.

4, in the present embodiment, the display unit 100 of the user terminal 10 displays domain regions in a tree structure to the user, and the user specifies the domain 45 to be selected in the designated domain display area. The domain can be selected by dragging and dropping (43) with a mouse or touch gesture (42). In addition, the designation can be released by dragging and dropping the domain 44 already selected in the designated domain display area 42 out of the area.

At this time, the domain tree may display or hide subdomains by the '+' button 46 or the '-' button 47.

Also, domains that are already selected among the nodes of the tree can be expressed differently so that the user can avoid unnecessary reselection. In addition, since the general area of the selected designated domain display portion 42 is selected in advance, it is displayed differently from other domains so that the user can recognize this fact.

Referring to FIG. 4, 'General', 'Touring', 'Restaurant', and 'Korean Food' are displayed differently from the unselected domains as domains for the designated domains that are already selected in the displayable area of the current display unit. To inform the user that it has been selected. The selected domain also appears in the designated domain display area 42, of which the 'General' domain 48 is selected by default, unlike other selected domains. .

FIG. 4 also shows that the current user drags 43 the 'Conference' domain 45 to the designated domain display area 42 in order to specify the conference related domain. Further, in the present embodiment, in addition to dragging and dropping a label of a domain for selecting a designated domain, it is also possible to select the designated domain by double-clicking or right-clicking a menu.

Furthermore, referring to FIG. 5, instead of a drag-and-drop method, the user selects the corresponding domain by checking a blank check box 51 through a user's click or touch. In addition, in the case of the general domain 53, which should always be selected, a check is displayed in a different color to inform the user that it is always selected.

6 illustrates an example of dynamically expressing a user interface for selecting a domain of the present invention. FIG. 6 expresses the tree structure of FIGS. 4 to 5 more dynamically, so that the user can operate it more easily. By clicking or touching the domain that the user wants to select, the child nodes are shown as sub-domains of the corresponding domain, and selection is possible (61). Also, domains of the designated domain that are not selected are represented differently and lost to the user.

7 illustrates an example of a user interface for designation of designated domains having an equivalent structure without a hierarchical structure. Each designated domain is represented by an icon. The label name of the above-described domain is displayed at the bottom of the icon to inform the user which domain the icon corresponds to. In addition, in the case of the icon form, all of them are displayed in the same form in FIG. 7, but it is also possible to configure the 'Medical' icon in an intuitive form such as '+' so as to inform the user of the corresponding domain.

The screen has an assignable domain display area 74 to inform the user what domain areas are selectable, and at the bottom of the screen there is a selected designated domain display area 71 that allows the user to select the currently selected and deselectable designated domains. Tell what it is. The user selects a designated domain by clicking or touching the domain 72 corresponding to the designated domain to be selected in the designable region display region or dragging and dropping the selected domain to the selected designated domain display region 71. The selected designated domain makes the icon disappear from the existing assignable area display area, thereby avoiding unnecessary reselection of the user. Similarly, the selected domain is clicked or touched or dragged and dropped to the designated domain display area to cancel the selection. The released designated domain appears in the assignable domain display area 74 to allow selection again. In addition, the scroll bar 75 is arranged to increase the accessibility of the user to easily access a large number of designated domains. In the present embodiment, the icons displayed on the assignable domain display area are changed by the scroll operation. However, in the case of the designated domain display area, it is preferable that the icons are not changed regardless of the scroll for selection through the icon.

FIG. 8 is an example of a user interface for designated domains having a hierarchical structure to which FIG. 7 is applied. When the user clicks or touches an icon as the domain 81 for the designated domain for which the user wants to view the designated subdomain, the icons for the subdomain of the corresponding domain appear in the subarea display portion. The lower area display area 82 shows a boundary so as to be distinguished from upper designated domain domains. In addition, the icons 84 may also be dragged and dropped to the designated domain display area 83 or selected as the designated domain by clicking or touching. In addition, when selecting / deselecting the upper designated domain in deselecting the designated domain, the lower designated domain included in the upper designated domain is automatically selected / released to support comprehensive selection / deselection.

In FIG. 8, when the user selects an icon 81 corresponding to the 'Seoul' domain, subdomains of the 'Seoul' domain, for example, 'Seoul Hotel' and 'Seoul Restaurant' are displayed. The user may select 'Seoul Hotel' 84 of the displayed icons or drag and drop 84 onto the designated domain display area 83 to select the Seoul hotel-related domain as the designated domain.

Furthermore, the display unit 100 according to the present embodiment may grasp the user situation using the user information collected through the user's terminal, and display the recommended domain to the user according to the discarded situation information. 9 shows an example of identifying a user's situation and presenting an appropriate designated domain.

The user information collected through the user's terminal is the user's location information through the GPS (Global Positioning System) built into the user terminal, the surrounding information through the camera, the ambient sound information recognized by the microphone, and the like. Figure out. Therefore, the speech recognition apparatus in the present embodiment recommends a domain that can be designated through the information on the user context to the user. For example, a user's GPS can recommend Seoul in Korea as a designated domain, and if a camera's surroundings are recognized as a restaurant through a camera, it is also possible to recommend travel and restaurants as a designated domain. You can also recommend an airport if you can hear the sound of takeoffs and landings through the microphone.

Therefore, in the present embodiment, it is preferable that the display unit 100 highlights domains of recommended designated domains and displays them on the screen. By helping the user to easily select only the designated domain from among the designated domains, it is possible to obtain a faster and more accurate speech recognition and automatic interpretation result by preventing the addition of recognition support data that supports the recognition of unnecessary vocabulary and sentence expressions. On the contrary, designated domains that are determined to be unnecessary or low in availability using contextual information are dimmed or not displayed, thereby facilitating easy recognition of the user and preventing unnecessary selection.

Referring to FIG. 10, when the user's location is recognized as Yeosu in Korea through current GPS information (101), an icon 102 (Yeosu Hotel, Yeosu Restaurant, Yeosu Expo) corresponding to the domain related to Yeosu is highlighted and displayed. The icon 103 (Medical) corresponding to the domain with low relevance is grayed out.

Furthermore, the domain display unit 100 according to the present embodiment may display at least one example recognition data for exemplifying a user to a voice recognition level designated according to the selection of a domain.

Referring to FIG. 11, when the user thinks of adding a 'Conference' related domain, and selects the 'Conference' domain as the designated domain through the recognition example section 114, the level of recognizable voice is “Where is the nearest Gal- bi buffet from the 8 th Advanced Computing (115) Conference hall? ”Can be indirectly illustrated to help users select a domain.

In the present embodiment, the user input unit 200 receives the selected domain from the user through the domain selection screen displayed on the display unit 100, and the communication unit 300 sends information on the selected domain to the voice recognition server 20. send.

The voice recognition server 20 performs voice recognition by referring to data corresponding to the domain selected by the user among the voice recognition reference data stored in the DB 30 through the received information on the selected domain. Then, the performed voice recognition result is transmitted to the user terminal 10.

The voice recognition server 20 according to the present embodiment described above is described as a system for communicating with the user terminal 10 and receiving a result from the voice recognition server for voice recognition. Accordingly, the voice recognition server 20 may be a voice recognition module inside the terminal, and the DB 30 may be implemented through an internal memory. In this case, the communication unit 300 of the user terminal 10 may provide information on the selected domain. The external voice recognition server 20 may be transmitted to the internal voice recognition module.

In this case, the configuration of the user terminal 10, the display unit 100, the user input unit 200, and the communication unit 300 operate as an interface module for selecting a voice recognition domain, and the voice recognition server interoperates with the voice recognition server. Implemented as a voice recognition module to perform the.

The voice recognition device 10 according to the embodiment of the present invention described above provides a user interface that can be easily understood and easily modified by a designated domain to users, thereby increasing the adaptability to a changing environment, thereby recognizing speech recognition. And can increase the accuracy of automatic interpretation. Hereinafter, an area designation method through the voice recognition apparatus 10 according to the present embodiment will be described.

Referring to FIG. 12, the voice recognition region designation method includes a domain selection screen display step S100, a domain selection input step S200, and a selection information transmission step S300.

In the domain selection screen display step (S100), the display unit 100 may display a screen for selecting a domain as a unit for a speech recognition region of a predetermined classification for a designated speech recognition of the speech recognition server 20 to a user. do.

In the domain selection input step (S200), the above-described user input unit 200 receives a selection of a domain from a user.

In the selection information transmission step (S300), the communication unit 300 described above transmits the user's selection information for the domain to the voice recognition server 20.

Detailed operations of each step of the area designation method are the same as those described above with respect to the display unit 100, the user input unit 200, and the communication unit 300, and description thereof will be omitted.

In the above example, the voice recognition operation has been described mainly. However, since the voice recognition operation is essential for the automatic interpretation, the same can be applied to the automatic interpretation. For example, the voice recognition server 10 of FIG. 1 may be an automatic interpretation server.

The various aspects, embodiments, implementations, or features of the present invention may be used individually or in any combination, and the various embodiments described herein may be implemented by, for example, software, hardware, or a combination thereof. have. According to a hardware implementation, the embodiments described herein include application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), and the like. The processor may be implemented using at least one of a processor, controllers, micro-controllers, microprocessors, and an electrical unit for performing a function. May be embodied as computer readable code on a computer readable medium is any data storage device that can store data that can later be read by a computer system Examples of computer readable media include read only memory, random access Memory, CD-ROM, DVD, magnetic tape, optical data storage. Readable medium can also, be stored and executed as computer-readable code is distributed manner, network-may be distributed across computer systems connected.

So far I looked at the center of the preferred embodiment for the present invention. It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. Therefore, the disclosed embodiments should be considered in an illustrative rather than a restrictive sense. The scope of the present invention is defined by the appended claims rather than by the foregoing description, and all differences within the scope of equivalents thereof should be construed as being included in the present invention.

Claims (18)

A display unit which displays a screen for selecting a domain for speech recognition to a user;
A user input unit for receiving a selection of a domain from the user; And
And a communication unit which transmits the user's selection information for the domain. The method of claim 1,
And the display unit displays a domain selectable by the user or a domain previously selected and deselectable by the user. The method of claim 1,
And the display unit classifies and displays domains representing the domains into hierarchies according to speech recognition levels. The method of claim 3, wherein
And the display unit displays a domain of the domain selected from the user among the displayed domains classified into hierarchies. The method of claim 3, wherein
And the hierarchical layer according to the speech recognition level is classified into a general region providing a basic speech recognition region according to the occurrence state of the voice, and the occurrence situation is reclassified according to the occurrence place. The method of claim 3, wherein
And the display unit displays a domain representing a domain corresponding to a lower layer of the selected domain according to the user's selection of the domain. The method of claim 1,
The display unit recognizes a user context using user information collected through the user's terminal, and displays a recommended domain to the user according to the discarded context information. The method of claim 1,
And the display unit displays at least one example recognition data to the user to illustrate the voice recognition level specified according to the selection of the domain to the user. A domain selection screen display step of displaying a screen for selecting a domain to a user as a unit for a speech recognition region of a predetermined classification for speech recognition;
A domain selection input step of receiving a selection of a domain from the user; And
And a selection information transmitting step of transmitting the selection information of the user for the domain. The method of claim 9,
The displaying of the domain selection screen may include displaying a domain selectable by the user or a domain previously selected and deselectable by the user. The method of claim 9,
The displaying of the domain selection screen may include classifying and displaying domains representing the domain into hierarchies according to speech recognition levels. The method of claim 11,
The displaying of the domain selection screen may include displaying a domain of the domain selected from the user among the displayed domains classified into hierarchies. The method of claim 11,
The hierarchical layer according to the voice recognition level is a hierarchical structure that classifies a general area providing a basic voice recognition area according to a voice occurrence situation, and reclassifies the occurrence situation according to a generation place. The method of claim 11,
The displaying of the domain selection screen may include displaying a domain representing a domain corresponding to a lower layer of the selected domain according to the user's selection of the domain. The method of claim 11,
In the displaying of the domain selection screen, the user situation is collected using the user information collected through the terminal of the user, and the recommended domain is displayed to the user according to the discarded situation information. The method of claim 11,
The displaying of the domain selection screen may include displaying, to the user, at least one example recognition data for illustrating to the user a speech recognition level designated according to the selection of the domain. A domain selection screen display step of displaying a screen for selecting a domain for speech recognition to a user;
A domain selection input step of receiving a selection of a domain from the user; And
A computer-readable recording medium having stored thereon a program for executing a selection information transmitting step of transmitting selection information of the user for the domain. An interface module for displaying a screen for selecting a domain for speech recognition to a user, receiving a selection of a domain from the user, and transmitting selection information of the user for the domain; And
And a speech recognition module configured to perform speech recognition by referring to data corresponding to a domain selected by the user among the reference data for speech recognition through the received selection information of the user.

Images