JP2012093508A - Voice recognition support system, voice recognition support device, user terminal, method and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce errors of voice recognition result of voice which is input at a place where a user inputs the voice for voice recognition while being out etc.SOLUTION: A voice recognition support system comprises: place recognition result accuracy information output means 501 for outputting place information indicating the place and information indicating accuracy of the voice recognition result on the place in association with each other; and voice recognition place fitness information output means 502 for outputting information on whether the place specified for inputting the voice to be recognized is fit for the voice recognition or not on the basis of the place information and the information indicating the accuracy of the voice recognition result output by the place recognition result accuracy information output means.

Description

  The present invention relates to a voice recognition support system, a voice recognition support device, a user terminal, a method, and a program that support a user's voice recognition by providing information on whether a certain place is suitable for voice recognition.

  A call center or the like recognizes a call, which is a response to an inquiry, by voice, and uses the resulting text to study CS improvement measures. In recent years, mobile devices such as mobile phones and smartphones have spread, and there is a trend to recognize voices spoken to mobile devices and use them. For example, after an outside salesperson talks about the contents of a business talk to a mobile device after the business talk, the voice-recognized text is stored on a server via the network, which leads to more efficient daily report creation after returning home.

  However, unlike indoors such as call centers, there are places where noise is constantly generated depending on the location outdoors, which is the main use place for mobile devices, and noise is mixed in the voice spoken in such places. Therefore, there is a problem that the result of speech recognition is adversely affected, that is, errors are increased. In addition, noise is not limited to the outdoors. For example, even indoors, there are few places that are supposed to be recorded in advance such as a call center, especially in places that have not been done so far. Since it is not known whether or not is suitable for speech recognition, there is a problem in that the speech remaining in the place is mixed with noise more than expected and the result of speech recognition is adversely affected.

  As one of countermeasures, there is a technique (for example, Patent Document 1) in which a device for removing noise is physically inserted in a portable device.

JP 09-252268 A

  However, the technique as described in Patent Document 1 is not incorporated in all portable devices and has low versatility.

  Therefore, the present invention provides a speech recognition support system and speech recognition that can reduce errors in speech recognition results of speech input at a place when the user performs speech recognition by inputting the speech while away from home. An object is to provide a support device, a user terminal, a method, and a program.

  The speech recognition support system according to the present invention includes location information indicating a location and information indicating the accuracy of a speech recognition result for the location in association with each other, and location information stored in association with the storage means. Speech recognition location suitability information output means for outputting information indicating whether or not a location designated as a location for inputting speech to be recognized is suitable for speech recognition based on the information indicating the accuracy of the speech recognition result. It is characterized by having.

  The speech recognition support apparatus according to the present invention stores location information indicating a location and information indicating accuracy of a speech recognition result for the location in association with each other, and is stored in association with the storage means. Based on the location information and the information indicating the accuracy of the speech recognition result, the speech recognition location suitability information output means for outputting information as to whether or not the location designated as the location for inputting the speech to be recognized is suitable for speech recognition. It is characterized by comprising.

  In addition, the user terminal according to the present invention associates location information with location information by transmitting location information to a predetermined server device when location information, which is a location for inputting recognition target speech, is input. Location recognition result accuracy information receiving means for receiving information indicating the accuracy of the speech recognition result for the location, location information, and information indicating the accuracy of the speech recognition result for the location received by the location recognition result accuracy information receiving means. And a voice recognition place suitability information output means for outputting information as to whether or not a place designated as a place for inputting the speech to be recognized is suitable for voice recognition.

  In addition, when the location information that is the location for inputting the speech to be recognized is input, the user terminal displays information indicating the accuracy of the speech recognition result for the speech input at the location indicated by the input location information. Based on the voice recognition result accuracy information collecting means to collect, the location information, and the information indicating the accuracy of the voice recognition result collected by the voice recognition result accuracy information collecting means, it is designated as the place to input the speech to be recognized. Voice recognition location suitability information output means for outputting information on whether or not the location is suitable for speech recognition may be provided.

  In addition, the speech recognition support method according to the present invention associates location information indicating a location with information indicating accuracy of a speech recognition result for the location in a predetermined storage device, and associates the information with the storage device. Based on the stored location information and information indicating the accuracy of the speech recognition result, information indicating whether or not the location designated as the location for inputting the speech to be recognized is suitable for speech recognition is output. And

  Further, the speech recognition support program according to the present invention includes a storage process in which a computer stores location information indicating a location and information indicating the accuracy of a speech recognition result for the location in a predetermined storage device, and a recognition target Execution of location recognition result accuracy information output processing for outputting information indicating the accuracy of the speech recognition result for the designated location stored in the storage device when location information that is assumed to be a location for inputting voice is input It is characterized by making it.

  Further, the voice recognition support program causes the computer to store the location information indicating the location and the information indicating the accuracy of the voice recognition result for the location in association with each other and store the information in a predetermined storage device, and the storage device Speech recognition that outputs information indicating whether or not the location specified as the location to which the speech to be recognized is input is suitable for speech recognition based on the location information stored and the information indicating the accuracy of the speech recognition result The place suitability information output process may be executed.

  Further, the speech recognition support program transmits location information to a predetermined server device and associates the location information with the location information when location information, which is a location for inputting speech to be recognized, is input to the computer. As a place to input the speech to be recognized based on the reception process for receiving information indicating the accuracy of the speech recognition result for the location, and the location information and the information indicating the accuracy of the speech recognition result for the received location A voice recognition place suitability information output process for outputting information on whether or not the designated place is suitable for voice recognition may be executed.

  In addition, the speech recognition support program, when the location information that is the location where the speech to be recognized is input is input to the computer, the accuracy of the speech recognition result for the speech input at the location indicated by the input location information Whether the location specified as the location where the speech to be recognized is input is suitable for speech recognition based on the collection process that collects information indicating the location and the location information and the information indicating the accuracy of the collected speech recognition results A voice recognition place suitability information output process for outputting information on whether or not may be executed.

  ADVANTAGE OF THE INVENTION According to this invention, when a user inputs a voice | voice in the place of going out and performs voice recognition, the error of the voice recognition result of the voice input at the place can be reduced.

It is a block diagram which shows the structural example of a speech recognition assistance system. It is a block diagram which shows the structural example of a speech recognition assistance server. It is a block diagram which shows the structural example of a user terminal. It is a flowchart which shows an example of the processing flow of a speech recognition process. It is a flowchart which shows an example of the processing flow of a recognition result text display process. It is a flowchart which shows an example of the processing flow of a recognition result text correction process. It is a flowchart which shows an example of the processing flow of a speech recognition conformity score calculation process. It is a flowchart which shows an example of the processing flow of a speech recognition conformity place judgment process. It is a flowchart which shows an example of the processing flow of the nearest optimal place presentation process. It is explanatory drawing which shows an example of the user table which user DB19 hold | maintains. It is explanatory drawing which shows an example of the speech recognition result table which speech recognition result DB17 hold | maintains. It is explanatory drawing which shows an example of the recognition place table which voice recognition conformity place DB18 hold | maintains. It is explanatory drawing which shows an example of the recognition result correction history table which voice recognition result DB17 hold | maintains. (A) is explanatory drawing which shows an example of a recognition result text list page, (b) is explanatory drawing which shows an example of a recognition result text correction page. It is explanatory drawing which shows the specific example of each step of a speech recognition conformity score calculation process. It is a block diagram which shows the outline | summary of this invention. It is a block diagram which shows the other structural example of the speech recognition assistance system 500 by this invention.

Embodiment 1. FIG.
Hereinafter, embodiments of the present invention will be described with reference to the drawings. First, an outline of the present embodiment will be described. The present embodiment provides information as to whether or not the place where the user wants to leave the voice, that is, the place where the voice to be recognized is input (recorded) is a place suitable for voice recognition. Further, when it is determined that the place where the speech to be recognized is input is not suitable for speech recognition, information on the nearest suitable place is provided. By providing such information, it is possible to prompt the user to move to the nearest suitable place, and as a result, it is possible to prevent noise from being mixed and reduce errors in the speech recognition result. Note that a place suitable for voice recognition refers to a place where a voice recognition result can be obtained with high accuracy when voice recognition is performed on voice input at that place. Note that a speech recognition result is obtained with high accuracy means that, for example, spoken speech is converted into accurate text and output.

  For example, a case is assumed in which a salesperson who goes around enters a voice outside and performs voice recognition processing on the voice. Possible causes of speech recognition errors include whether the spoken word is present in the dictionary or whether it is a clear sound, but if it is a clear sound that does not contain noise, it has a significant effect on the accuracy of speech recognition. As a matter of course, there are fewer errors in speech recognition in an environment where noise is not mixed. That is, it can be considered that as the noise level estimated at the time of the voice recognition process or the amount of correction performed on the voice recognition result by the user increases, the voice is left in a place that is not suitable for voice recognition.

  FIG. 1 is a block diagram illustrating a configuration example of a voice recognition support system according to the present embodiment. As shown in FIG. 1, the speech recognition support system of this embodiment includes a speech recognition support server 100 and one or more user terminals (in this example, user terminals 301, 302, and 303). It is assumed that the voice recognition support server 100 and each user terminal are connected via the network 200, respectively.

  FIG. 2 is a block diagram illustrating a configuration example of the voice recognition support server 100. As shown in FIG. 2, the speech recognition support server 100 includes a speech recognition unit 11, a recognition result text display unit 12, a recognition result text correction unit 13, a speech recognition suitability score calculation unit 14, and a speech recognition suitability place determination. A unit 15, a nearest optimum place presentation unit 16, a speech recognition result database (DB) 17, a speech recognition compatible place database (DB) 18, and a user database (DB) 19 are provided.

  The voice recognition unit 11 receives voice data from the user terminal and executes voice recognition. The recognition result text display unit 12 displays the text of the result of voice recognition on the user terminal. The recognition result text correction unit 13 corrects the text resulting from the speech recognition.

  The voice recognition suitability score calculation unit 14 calculates the degree of suitability for voice recognition of the place in association with the place where the voice is input. In the present embodiment, the speech recognition suitability score calculator 14 calculates a speech recognition suitability score indicating the degree of suitability (satisfaction level) for speech recognition at the location.

  The speech recognition conformity location determination unit 15 determines whether the location is suitable for speech recognition, more specifically, from the speech recognition conformity score associated with the location where the user is about to input (record) speech. Determine if it is suitable as an input location. As a result of the determination, the nearest optimum place presenting unit 16 presents the place most suitable for speech recognition near the place.

  The voice recognition result DB 17 stores voice recognition result data. The speech recognition result DB 17 may hold, for example, speech recognition results performed on speech input by the user of the system at various locations in association with information for identifying the speech recognition results. . Further, the voice recognition result DB 17 holds information indicating the correction contents of the voice recognition result text by the user, together with the information of the voice recognition result.

  The voice recognition compatible place DB 18 holds information indicating the degree of suitability for voice recognition of the place where the voice is left. For example, the voice recognition suitable place DB 18 may hold information indicating a place and a voice recognition suitability score for the place in association with each other.

  The user DB 19 holds user information that is information about the user. For example, the user DB 19 may hold information for identifying the user and information indicating the accuracy of the voice recognition result for the user in association with each other. Note that the information indicating the accuracy of the speech recognition result for the user is information used for calculating the speech recognition conformity score of the place so that the influence of the user characteristics is not reflected.

  FIG. 3 is a block diagram illustrating a configuration example of the user terminal 301. As shown in FIG. 3, the user terminal 301 includes a transmission / reception unit 31, a voice input unit 32, a display unit 33, and a position information input unit 34.

  The transmission / reception unit 31 transmits / receives various information to / from the voice recognition support server 100 via the network 200. The voice input unit 32 inputs voice (voice data) that is a voice recognition target. The voice input unit 32 includes a microphone and inputs voice data collected by the microphone in response to a user operation. The display unit 33 displays a screen for allowing the user to input information received from the voice recognition support server 100 or information to be transmitted to the voice recognition support server 100. The position information input unit 34 inputs position information indicating a place where the voice is input and position information of a place where the conformity place determination is requested. The position information input unit 34 includes, for example, a GPS receiver, and inputs position information indicating a place where the voice is input by reading latitude / longitude information held by the GPS receiver when the voice is input. Also good. Further, for example, information designated (input operation) by a user may be input via an information input device such as a keyboard or a mouse. Information specified by the user (input operation) may be used as the position information indicating the place where the voice is input, instead of the position information of the place where the conformity place determination is requested.

  The user terminal performs transmission / reception and display of information, input of audio data, and the like according to a control instruction from a CPU that operates according to a program, for example. This program is not limited to a program stored in the storage device in advance, but includes a program downloaded by accessing the server using a browser or the like.

  Moreover, although the example of the user terminal 301 was shown in FIG. 3, other user terminals (user terminal 302, 303, etc.) may be the same. The operation of this system can be divided into several phases such as from voice input to voice recognition, confirmation and correction of the voice recognition result, and suitability determination of the voice recognition place. Therefore, it is possible to use different user terminals, for example, to divide the user terminals that execute each phase according to the user's usage mode.

  In the present embodiment, the speech recognition unit 11, the recognition result text display unit 12, the recognition result text correction unit 13, the speech recognition conformity score calculation unit 14, the speech recognition conformity place determination unit 15, and the nearest optimum place presentation unit 16 For example, it is realized by a data processing device provided in the voice recognition support server. The voice recognition result DB 17, the voice recognition matching place DB 18, and the user DB 19 are realized by, for example, a memory and a database system provided in the voice recognition support server. The database system includes a control unit that outputs a group of data that matches the conditions as a series of data in response to a request.

Next, the operation of this embodiment will be described. In the present embodiment, the following processing flow is roughly included.
(1) When voice recognition is performed, location information and recognition results are stored in the system.
(2) When the voice recognition result is corrected, the correction content is stored in the system.
(3) From the difference between the correction amount of the recognition result and the average correction amount of the individual, a score is calculated as to whether the location is suitable for speech recognition.
(4) When the user is going to input a voice to be recognized in a place where the score is low, the user presents that the place is not suitable for voice recognition, and presents a place where the nearest score is high.

  In this way, by making it available to an unspecified number of users via the network, the speech recognition results of a wide range of users are accumulated daily as usage histories, and score accuracy is improved based on the results. As a result, it is possible to obtain information as to whether or not the place is suitable for voice recognition even in a place where a certain user has never been.

  Next, each processing flow will be described in more detail with reference to FIGS. First, a process for recognizing voice input using a user terminal will be described with reference to FIG. FIG. 4 is a flowchart illustrating an example of a processing flow of the voice recognition processing. In this example, a mobile phone is assumed as the user terminal.

  As shown in FIG. 4, first, a voice is recorded using a user terminal (S1001). In this example, it is assumed that the voice is recorded in the system by calling the voice from the mobile phone to the specific telephone number held by the system. For example, the user terminal inputs voice in response to an operation from the user, and outputs (transmits) the input voice (voice data) to the voice recognition unit 11 of the voice recognition support server 100. Note that the voice recording method is not limited to this.

  Next, when receiving the recognition target voice, the voice recognition unit 11 acquires latitude / longitude information, date / time, and mobile phone number (information that can identify the user) of the place where the voice is input (S1002). In this example, it is assumed that the system (more specifically, the voice recognition unit 11) acquires the latitude / longitude information held by the GPS receiver built in the user terminal via the network. . In addition, it is assumed that the date and time and the mobile phone number are obtained at the time of voice recording. In addition, the method is not limited to the method of acquiring from the system side, the user terminal voluntarily reads out at the timing when the voice recording starts and transmits the voice data together with the latitude / longitude information of the current position, the current date and time, and the mobile phone. You may make it transmit a telephone number.

  Further, the voice recognition unit 11 acquires a user ID having a telephone number that matches the acquired mobile phone number from the user table (see FIG. 10) of the user DB 19 (S1003).

  FIG. 10 is an explanatory diagram showing an example of a user table held by the user DB 19. The user table shown in FIG. 10 is a table for registering user information, and includes a user ID, a user's first and last name, a telephone number, a voice recognition score (average), and the number of registrations.

  The user ID is an identifier for identifying the user in the system. The voice recognition score is information indicating the accuracy of the voice recognition result for the user. In this embodiment, the voice recognition score (average) calculated by the voice recognition matching score calculation unit 14 for each user is registered. To do. The number of registrations is the number of times the user has registered the speech recognition result in this system. Note that the user ID, first and last name, and telephone number are registered in advance before the system starts the processing. A method for registering the voice recognition score for each user and updating the number of registrations will be described later.

  Next, the voice recognition unit 11 performs voice recognition on the voice recorded in S1001 (S1004). In this example, it is assumed that a known speech recognition apparatus as described in JP-A-2002-099296 is installed in the system as a speech recognition method.

  When the voice recognition is completed, the voice recognition unit 11 uses the user ID, the recognition result text obtained by performing the voice recognition in S1004, the latitude / longitude, and the date / time on the voice recognition result table of the voice recognition result DB 17 (see FIG. 11). (S1005).

  FIG. 11 is an explanatory diagram showing an example of a speech recognition result table held in the speech recognition result DB 17. The speech recognition result table shown in FIG. 11 is a table for registering speech recognition result data, and includes a recognition result ID, a user ID, recognition result text, date information, and location information.

  The recognition result ID is an identifier for identifying one recognition result in the system. The user ID is information for identifying the user who has input the voice obtained from the recognition result. In this example, the user ID of the user table is registered. The recognition result text is a text indicating a recognition result obtained by voice recognition. The date and time information is information indicating the date and time when the voice obtained from the recognition result is input (recorded). The location information is information (latitude and longitude in this example) that indicates the location where the voice obtained from the recognition result is input (recorded).

  For example, the user “Kazuo Suzuki” recognized the voice left at the latitude “35.651386” and longitude “139.754183” at 15:10 on July 12, 2010 using a mobile phone. In this case, data such as the record 111 in FIG. 11 is registered, and the recognition result ID (34 in this example) is numbered and stored.

  Next, the speech recognition processing unit 11 has location information that matches the location (latitude / longitude) where the speech that has been speech-recognized this time is input in the recognition location table (see FIG. 12) of the speech recognition compatible location DB 18. It is determined whether or not to perform (S1006). If it does not exist (No in S1006), this latitude / longitude is registered as a new voice recognition place in the recognition place table of the voice recognition compatible place DB 18 (S1007). Then, the voice recognition process ends. On the other hand, if it exists (Yes in S1006), the process is terminated as it is.

  FIG. 12 is an explanatory diagram showing an example of a recognition location table held by the speech recognition compatible location DB 18. The recognition place table shown in FIG. 12 is a table for registering the voice recognition suitability score for each voice recognition place, and includes a recognition place ID, place information, a voice recognition suitability score, and the number of registrations.

  The recognition place ID is an identifier for identifying one recognition place in the system. The location information is information indicating the location. The speech recognition suitability score is information indicating the accuracy of the speech recognition result for the place. In this embodiment, the speech recognition suitability score calculated by the speech recognition suitability score calculation unit 14 is registered. The number of registrations is the number of times the place is registered as a recognition place in the system. Note that the voice recognition fitness score and the number of registrations may be empty data (no voice recognition score registered, number of registrations 0, etc.) when the voice recognition unit 11 registers a new recognition place in S1007. A method for registering the speech recognition fitness score and updating the number of registrations will be described later.

  Next, with reference to FIG. 5, a recognition result text display process for displaying the voice-recognized result text to the user will be described. FIG. 5 is a flowchart illustrating an example of a processing flow of recognition result text display processing. In this example, it is assumed that the user holds a terminal such as a personal computer, and the system creates a screen for the Web browser installed in the terminal, and displays information using the screen. Assumes to do. In addition, the terminal and method to display are not this limitation. Further, it is assumed that the user transmits a user ID to the system at the first time of referring to the screen with a Web browser, and personal authentication is performed.

  In the following description, it is assumed that “Kazuo Suzuki” of the user ID “1” is referring. Here, the user terminal accesses a predetermined Web page prepared by the system in response to a request from the user. Therefore, when a user ID is input, a recognition result text display request is transmitted to the recognition result text display unit 12 of the speech recognition support server 100 together with the user ID.

  As shown in FIG. 5, when receiving the recognition result text display request, the recognition result text display unit 12 matches the received user ID from the speech recognition result table (see FIG. 11) of the speech recognition result DB 17. Speech recognition result data (more specifically, recognition result text) is acquired (S2001). Then, the list of acquired recognition result texts is displayed on the screen of the user terminal that is the request source (S2002).

  FIG. 14A is an explanatory diagram illustrating an example of a recognition result text list page. As shown in FIG. 14A, the speech recognition result text for the speech registered by the user may be sequentially displayed together with the recognition result ID. In the example shown in FIG. 14A, an operation button (correction button) for giving an instruction to start the correction operation of the voice recognition result text is provided for each voice recognition result text display area. Yes.

  When the user confirms and edits one recognition result from such a recognition result list, the user may press a correction button as shown in FIG. At this time, the user terminal transmits the start of correction and the recognition result ID to the voice recognition support server 100 according to the operation. In this example, “34” is transmitted to the voice recognition support server 100 as the recognition result ID.

  The recognition result text display unit 12 acquires the recognition result text that matches the received recognition result ID from the speech recognition result table of the speech recognition result DB 17 (S2003), and displays the recognition result text correction page (S2004).

  FIG. 14B is an explanatory diagram illustrating an example of a recognition result text correction page. As shown in FIG. 14 (b), the recognition result text to be corrected may be displayed together with the recognition result ID. The recognition result text on the recognition result text correction page is displayed so that the user can edit it. In the example shown in FIG. 14B, an operation button (correction confirmation button) for instructing confirmation of the correction content of the voice recognition result text is provided.

  Next, with reference to FIG. 6, the recognition result text correction in which the user corrects the recognition result text will be described. FIG. 6 is a flowchart illustrating an example of a processing flow of recognition result text correction processing.

  Now, it is assumed that the recognition result text correction page is displayed on the user terminal in S2004 of FIG. 5 (S3001). Note that the recognition result text correction unit 13 may receive the start of correction and the recognition result ID, and display the recognition result text correction page in the recognition result text correction processing.

  When the user corrects the error in the recognition result text and presses the correction confirmation button, the fact is transmitted to the recognition result text correction unit 13. When the recognition result text correction unit 13 receives the message indicating the correction confirmation, the recognition result text correction unit 13 acquires the recognition result ID and the corrected recognition result text, and together with the current date and time, the recognition result correction history table of the voice recognition result DB 17 (see FIG. 13). (S3002).

  FIG. 13 is an explanatory diagram showing an example of a recognition result correction history table held in the voice recognition result DB 17. The recognition result correction history table shown in FIG. 13 is a table for registering the correction contents of the speech recognition result text by the user, the correction history ID, the recognition result ID, the (corrected) recognition result text, and the date and time. Information.

  The correction history ID is an identifier for identifying the correction content. The recognition result ID is an identifier for identifying the recognition result text to be corrected. As the recognition result ID, the recognition result ID of the recognition result table is registered. The recognition result text is the recognition result text after being corrected by the user. The date / time information is information indicating the corrected date / time.

  For example, when the recognition result text is corrected to the contents shown in FIG. 14B, data such as the record 131 in FIG. 13 is registered in the recognition result correction history table, and the correction history ID (100 in this example) is registered. ) Is numbered and stored.

  Next, with reference to FIG. 7 and FIG. 15, a speech recognition suitability score calculation process for calculating a speech recognition suitability score indicating the degree of suitability for speech recognition at a place where speech is input will be described. FIG. 7 is a flowchart illustrating an example of a processing flow of the speech recognition suitability score calculation process. FIG. 15 is an explanatory diagram showing a specific example of each step shown in FIG.

  Hereinafter, as illustrated in FIG. 14, a case where the user ID “1” “Kazuo Suzuki” modifies the recognition result ID “34” will be described as an example. For example, when the corrected recognition result text is registered in the recognition result correction history table, the recognition result text correction unit 13 notifies the voice recognition suitability score calculation unit 14 of that fact. In the example illustrated in FIG. 7, the voice recognition suitability score calculation unit 14 that has received the notification firstly recognizes the recognition result text that matches the recognition result ID of the corrected recognition result text (the recognition result text before correction), the location information, Is obtained from the speech recognition result table of the speech recognition result DB 17 (S4001).

  Next, the correction result text that matches the recognition result ID is acquired from the recognition result correction history table of the voice recognition result DB 17 (S4002). Here, the speech recognition text before correction is read out.

  Next, as the amount of correction, the number of different characters when the recognition result text and the correction result text are compared is counted, and the ratio of the number of different characters in the number of recognition result text is calculated (S4003). In this example, the number of characters in the recognition result text is “45”, and the number of different characters (correction amount) in the recognition result text and the correction result text is “14”. /45=0.31 ”(see S4003 in FIG. 15). In this example, the ratio of the number of different characters in the number of recognition result texts calculated in step S4003 is treated as a speech recognition score indicating the accuracy of speech recognition for the speech recognition result. It should be noted that the voice recognition score of this example has a property that the value increases as the ratio of the number of different characters increases, that is, as the voice recognition accuracy decreases.

  Next, the user ID that matches the recognition result ID is acquired from the voice recognition result table of the voice recognition result DB 17 (S4004). Here, the user ID that registered the voice recognition result is read out.

  Next, the voice recognition score (average) matching the user ID is acquired from the user table (see FIG. 10) of the user DB 18, and the difference from the current voice recognition score calculated in S4003 is calculated (S4005). ). Here, it is calculated how much the accuracy of the current speech recognition result differs from the accuracy of the average speech recognition result of the user. In this example, as shown in the record 101 in FIG. 10, a voice recognition score (average) “0.21” is acquired, and the difference is “0.31−0.21 = 0.1” (in FIG. 15). (See S4005). This value means that it is determined that the voice recognition place is a place that deteriorates the 0.1 score.

  In the present embodiment, the difference obtained in step S4005 is referred to as a speech recognition suitability score. Strictly speaking, the speech recognition suitability score obtained in step S4005 is a speech recognition suitability score indicating the accuracy of the speech recognition result at that location based on the current speech recognition result. Note that the speech recognition score (average) obtained from the user table is the speech recognition score indicating the accuracy of the results of speech recognition by the user at various locations so far (here, the speech recognition text and its correction result text). This is a numerical value obtained by averaging the ratios of different numbers of characters), and this numerical value serves as an index indicating the suitability of the individual user for speech recognition. The voice recognition fitness score calculated in S4005 is the difference between the voice recognition score obtained as a result of voice recognition performed on the voice recorded at a certain location this time, and the voice recognition score (average) of the user. When is larger than 0, it can be considered that the current place has a lower degree of fitness for speech recognition than the average.

  Next, based on the current speech recognition score calculated in S4003, the user's speech recognition score (average) is recalculated and registered in the speech recognition score (average) in the user table of the user DB 10 (S4006). ). In this case, since the previous voice recognition score (average) is “0.21”, the number of registrations is “15”, and the current voice recognition score is “0.31”, the number of registrations is one. It increases and changes to “(0.21 * 15 + 0.31) /16=0.22 (rounded to the second decimal place)” (see S4006 in FIG. 15).

  Next, the recognition location ID and the speech recognition fitness score (average) of the location where the latitude / longitude acquired in S4001 match are acquired from the recognition location table (see FIG. 12) of the speech recognition compatible location DB 18 (S4007). Here, the voice recognition conformity score (average) at the current location is read out. In this example, as shown in the record 121 of FIG. 12, the recognition location ID “5” and the speech recognition conformity score (average) “0.12” are acquired.

  Here, the speech recognition suitability score (average) at the recognition location is a numerical value obtained by calculating the average of the speech recognition suitability scores when various users recognize speech at this location. This numerical value serves as an index indicating the degree of fitness for speech recognition at the location, and the larger the value, the lower the degree of fitness for speech recognition. Note that, similar to the speech recognition suitability score, when this value is larger than 0, it can be considered that a correction amount larger than the average correction amount of the system user is expected. On the other hand, if the value is negative, it can be considered that the correction amount is smaller than the average correction amount of the system user.

  Next, based on the speech recognition suitability score calculated in S4005, the speech recognition suitability score (average) at that location is recalculated and registered in the speech recognition suitability score (average) in the recognition location table of the speech recognition suitability location DB18. (S4008). The method for obtaining the average may be a simple average or a square average. In this case, since the speech recognition conformity score (average) of the place so far is “0.12” and the number of registrations is “3”, the number of registrations is increased by “(0.12 * 3 + 0.1) /4=0.12 (rounded to the second decimal place) ”. Here, for example, weighting may be performed when the speech recognition suitability score calculated this time is added to the speech recognition suitability score (average) at the location. For example, if it is known in advance that the current voice recognition conditions are not preferable as a voice input location, the contribution may be lowered by applying a low weighting factor.

  Next, referring to FIG. 8, when a user tries to leave a voice using a mobile phone, a voice recognition suitable place determination process for determining whether or not a place where voice input is performed is suitable for voice recognition. explain. FIG. 8 is a flowchart illustrating an example of a processing flow of the voice recognition conformity place determination process.

  First, the voice recognition conformity place determination unit 15 of the voice recognition support server 100 acquires the latitude / longitude of the place where the user wants to leave the voice from the mobile phone (S5001). The acquisition method may be the same as the method assumed by the voice recognition unit 11. The location information is not necessarily location information indicating the current position. For example, the user may be allowed to input the place where the voice is about to be input through the user terminal.

  Next, the recognition location ID of the location where the acquired latitude / longitude matches is acquired from the recognition location table (see FIG. 12) of the speech recognition compatible location DB 18 (S5002), and it is determined whether the corresponding recognition location ID exists. (S5003). If the corresponding recognition place ID does not exist (No in S5003), the voice recognition unit 11 can perform voice recognition by recording the voice as it is and determining that the voice recognition suitability score has not yet been calculated. That's fine.

  In the following description, it is assumed that the place with the recognized place ID “7” indicated by the record 122 in FIG. When the corresponding recognition place ID exists (Yes in S5003), the speech recognition conformity place determination unit 15 calculates the speech recognition conformity score (average) of the places where the recognition place IDs match from the recognition place table of the speech recognition conformity place DB18. Obtain (S5004). In this example, “0.43” is acquired.

  Next, it is determined whether or not the acquired voice recognition conformity score (average) is a numerical value equal to or greater than a predetermined threshold (X) (S5005). Here, X is an arbitrary numerical value of 0 or more and 1 or less, and is a numerical value that is determined by the system administrator that the voice recognition conformity score (average) is equal to or greater than this numerical value and that the place is not suitable for speech recognition. The administrator can set and change freely. As a result of the determination, if the speech recognition suitability score (average) is equal to or greater than the predetermined threshold (X) (Yes in S5005), the process proceeds to the nearest optimum place presentation process. In this example, it is assumed that X is set to 0.3. Then, since the voice recognition conformity score (average) “0.43” is larger, the process proceeds to the nearest optimum place presentation process. On the other hand, if the voice recognition conformity score (average) is smaller (No in S5005), the voice recognition unit 11 may perform voice recognition by recording the voice as it is.

  Next, with reference to FIG. 9, the nearest optimum place presentation process for presenting the optimum place for speech recognition closest to the user will be described. FIG. 9 is a flowchart illustrating an example of the processing flow of the nearest optimal place presentation processing.

  In the nearest optimum place presentation process, the nearest optimum place presentation unit 16 is close to the place and the voice recognition suitability score (average) is based on the longitude / latitude information acquired in S5005 of the voice recognition suitability place determination process. A place lower than X is searched from the recognition place table of the voice recognition compatible place DB 18 to obtain a maximum of Y places (S6001). Here, Y is an arbitrary integer of 1 or more, and can be set and changed freely by the system administrator. In this example, it is assumed that Y is set to 1, and it is assumed that the location of the recognition location ID “12” shown in the record 123 of FIG.

  When the location information that matches the conditions is acquired, the nearest optimum location presentation unit 16 displays the location information on the screen of the requesting mobile phone. In this example, the nearest map is displayed, and the latitude / longitude location acquired in S6001 is plotted and displayed (S6002). In addition, when there is no information on a place that matches the condition, the fact is notified, or the distance from the acquired place is within a predetermined range and the voice recognition conformity score ( If there is a place where the average) is low, that is, the fitness is higher, information on the place may be presented. In this example, the case where a list of a plurality of places is shown on the map has been described. However, one place may be narrowed down and presented. For example, it may be narrowed down to a place where the voice recognition suitability score is the best within a predetermined range, or a place where the distance from the most designated place is closest among those where the voice recognition suitability score satisfies the threshold may be narrowed down.

  As described above, according to the present embodiment, based on the degree of suitability for speech recognition of a place where speech is to be left, if the suitability is low, the nearest place with the best suitability is presented to promote movement. Thus, errors in the speech recognition result can be reduced.

  Because, based on the information indicating the accuracy of each speech recognition result (in this example, the amount of correction of the speech recognition result), the speech recognition conformity score (average) of the speech recognition location is calculated, and the speech recognition conformity score (average) This is because it is determined whether or not the location is suitable for speech recognition. Note that, according to the present embodiment, regardless of whether the place where the sound is to be left is outdoors or indoors, even if the user has never actually visited the place. It is possible to determine whether or not the place is suitable for voice recognition.

  Further, in the present embodiment, the speech recognition conformity score (average) is calculated so that the influence due to the individual characteristics is difficult to be reflected. For example, even if multiple people speak the same sentence, the voice recognition result differs depending on the person because the loudness, pronunciation, and illness are different. For this reason, it is difficult to correctly calculate the suitability for a specific place simply by calculating the speech recognition score based on the correction amount of the speech recognition text. This is because it is impossible to determine whether the voice recognition is good or bad depending on the place or the person.

  In the present embodiment, the difference between the average of the past correction amount of the individual and the correction amount at the location is calculated, and the result value is used for calculating the speech recognition conformity score for the location. Incorporating processing to prevent the bias. For this reason, the place suitable for the nearest speech recognition can be presented with higher accuracy.

  In addition, since the calculation of the speech recognition suitability score (average) of the speech recognition place is repeated every day using the speech recognition of various users and the correction log of the recognition result, the accuracy of the suitability determination can be improved.

  In the above embodiment, an example of registering speech recognition text as information indicating the speech recognition result in the speech recognition result table has been shown, but not only the speech recognition result text but also a noise level that affects speech recognition, Information indicating the accuracy of the speech recognition result obtained as a result of speech recognition, such as the likelihood of the speech recognition result text, may be included.

  In the above embodiment, an example in which latitude / longitude is registered as location information has been shown. However, in addition to latitude / longitude, information indicating what the location is is classified into information such as information previously classified. Information indicating the characteristics of the location affecting the recognition result may be included. For example, the speech recognition suitability score calculation unit 14 may average these noise levels and likelihoods for each place and output them as a speech recognition suitability score (average). In such a case, the speech recognition conformity location determination unit 15 determines whether or not the average noise level or likelihood for each location that is output satisfies a threshold value, and whether or not it is suitable for speech recognition. What is necessary is just to judge. As in the case of the correction amount, the speech recognition fitness score for the input location of the speech is calculated using the difference between the current noise level and likelihood and the average of the user's past noise level and likelihood. May be calculated.

  In the above embodiment, an example in which the unit of one place is determined based on whether or not the latitude and longitude coincide with each other. However, the site indicated by a certain standard (within radius xm, number of floors, map, etc.) It is also possible to use units (the range of locations in such a classification) that are considered to be the same location along the inside. In such a case, it may be determined whether or not the location indicated by the acquired location information belongs within a certain range from the location indicated by the location information registered in the recognized location table.

  In the above embodiment, an example is shown in which a speech recognition suitability score is calculated as information indicating the accuracy of the speech recognition result for the place. However, information indicating the accuracy of the speech recognition result for the location is the speech recognition suitability score. Not limited. For example, the voice recognition score for each voice may be classified and averaged for each voice input place. Further, for example, the noise level included in the voice detected by voice recognition processing or the like for the voice input at the place can be used as it is. In such a case, instead of the processing of the speech recognition suitability score calculation unit 14, the noise level detected by the speech recognition processing is used as the location information of the place where the speech is input as information indicating the accuracy of the speech recognition result. Can be stored in association with each other.

  In the above embodiment, when voice input is started, it is determined whether or not the place is suitable for voice recognition, and when it is determined that the place is not suitable, the nearest suitable place is presented. The output example of information on whether or not the place is suitable is not limited to this. For example, when a location that is a candidate for a voice input location is designated, information indicating the accuracy of the speech recognition result for that location (speech recognition suitability score or noise level) is output as it is, so that the user can determine whether or not it is appropriate. It is also possible to judge. Further, for example, it is possible to output the evaluation value in a stepwise manner (highly suitable / suitable for a middle step / not so suitable) based on the information.

  In the above embodiment, the voice recognition support server 100 includes the voice recognition unit 11. However, the voice recognition unit 11 may be an external system. In such a case, instead of the voice recognition unit 11, a collection unit that collects the voice recognition result data from the external system may be provided.

  In the above embodiment, an example is shown in which the process from the voice recognition process to the nearest optimum place presentation process is executed on the voice recognition support server 100 side. However, these processes may be executed by the user terminal alone. Is possible. In such a case, a speech recognition support device or a speech recognition support program may be incorporated into the user terminal. That is, the user terminal includes a speech recognition unit 11, a recognition result text display unit 12, a recognition result text correction unit 13, a speech recognition conformity score calculation unit 14, a speech recognition conformity place determination unit 15, and a nearest optimum place presentation unit 16. May be provided. Further, the voice recognition unit 11, the recognition result text display unit 12, and the recognition result text correction unit 13 are omitted, and the present system includes a voice input location as information indicating accuracy of each voice recognition result and information of the voice. It is also possible to implement from the process of collecting information. Further, some of these may be provided on the server side and the other part on the terminal side.

  For example, when location information, which is a location for inputting speech to be recognized, is input on the user terminal side, the location information is transmitted to the server, and the speech recognition conformity associated with the location information is transmitted. A score or the like may be received, and information on whether or not the location is suitable for voice recognition may be output from the received information. In such a case, on the user terminal side, a receiving unit that receives information indicating the accuracy of the speech recognition result for the location specified from the server, the speech recognition conformity location determination unit 15, and the nearest optimum location presentation What is necessary is just to provide the part 16. FIG.

  Also, for example, when location information that is a location for inputting speech to be recognized is input on the user terminal side, the accuracy of the speech recognition result for each speech input at the location indicated by the input location information Information (noise level, correction amount, etc.) is collected, and information indicating whether or not the location specified as the location where the speech to be recognized is input from the collected information is suitable for speech recognition may be output. Good. In such a case, on the user terminal side, a collection unit that collects information indicating the accuracy of the speech recognition result for each voice input at the location indicated by the input location information, and the location from the collected information What is necessary is just to provide the production | generation part which produces | generates the information which shows the precision of the speech recognition result about, the speech recognition suitable place judgment part 15, and the nearest optimal place presentation part 16. In addition, you may make it provide the speech recognition suitable score calculation part 14 as a production | generation part which produces | generates the information which shows the precision of the speech recognition result about the place from the collected information.

  The user terminal includes a voice recognition result DB 17, a voice recognition place DB 18, and a user DB 19 that store data for one user without collecting voice recognition result data or the like via the network. But it is feasible. Further, some of these may be provided on the server side and the other part on the terminal side.

  In the above embodiment, at the timing when the correction of the voice recognition result is confirmed, the voice recognition suitability score is calculated and updated, the value is stored in the storage means, and the place where the voice is left from the user is stored. The example of checking the speech recognition conformity score held in the storage means when determining the suitability determination request and determining whether or not it conforms is shown, but the data retained in the storage means However, the content and the timing of score calculation are not limited to this. For example, each speech recognition result and its correction result are stored in a database, and a necessary result is read based on location information specified at the time of a judgment request from the user, and a speech recognition conformity score for that location is read. It is also possible to calculate and determine suitability. Further, for example, the database stores information indicating the suitability determination result together with the voice recognition suitability score calculated at the time of confirming the change, and only reads the stored suitability determination result when a determination request is made from the user. Thus, it is possible to present a response or the nearest optimal place based on the result.

Embodiment 2. FIG.
In this embodiment, information on the date and time when voice recognition is performed is used. Ambient noise in a specific place changes depending on the date and time. For example, the same place is noisy on weekends but quiet on weekdays, noisy in summer but quiet in other seasons.

  In the present embodiment, such features are reflected in the speech recognition suitability score based on the accumulated speech input date / time information. That is, by calculating the characteristics of the voice recognition conformity score at a specific location from the stored voice recognition date and time information and holding it as data, the voice recognition score that meets the date and time conditions for voice recognition by the user To return.

  More specifically, the speech recognition score can be registered in the speech recognition result table according to the first embodiment so that a speech recognition score for each predetermined date and time classification (for each time zone, for each season, for each day of the week, etc.) can be registered. Add a column. When the speech recognition score is registered in the speech recognition result table, the speech recognition score is registered in a speech recognition score column of a classification classified based on date / time information on which the speech is input. Similarly, a speech recognition suitability score for each predetermined date and time category can be registered in the recognition place table. For example, the date and time, the time zone, and the speech recognition conformity score are stored in association with each other. Further, when obtaining the speech recognition conformity score, it is calculated using only the speech recognition scores of the same category.

  For example, it is assumed that a speech recognition fitness score for each time zone is calculated. For example, the time zone is defined as follows. Midnight: 0-4 o'clock, early morning: 4-6 o'clock, morning: 8-12 o'clock, daytime: 12-16 o'clock, evening: 16-19 o'clock, night: 19-24 o'clock. It is possible to provide more detailed information by arbitrarily defining the season and day of the week.

  First, only information on a row corresponding to a specific place is acquired from the speech recognition result table. It is determined which time zone the time of the corresponding row matches. The calculated speech recognition suitability score is registered in the recognition place table together with the place information, date / time information, and time zone information. At this time, if a speech recognition suitability score (average) in which the location information and the time zone information match is already registered, the average is recalculated taking into account the speech recognition suitability score calculated this time. Furthermore, a speech recognition suitability score that does not consider the time zone is also calculated.

  In the speech recognition suitability place determination process, the speech recognition suitability score (average) in the same time zone of the recognition place table and the speech recognition suitability score (average) not considering the time zone, that is, the place obtained by the same method as in the first embodiment. If the former is higher, the location is not suitable for recognition in the specified time zone, and the voice recognition suitability score is the same for the same time zone. Present the lowest nearest location.

  It is also possible to determine suitability by comparing the speech recognition suitability score (average) in the same time zone of the recognition location table with a predetermined threshold (X).

  As described above, according to the present embodiment, more accurate information can be provided. Therefore, when a user performs voice recognition by inputting voice while away from home, the voice input at the place Errors in speech recognition results can be further reduced.

  Next, the outline of the present invention will be described. FIG. 16 is a block diagram showing an outline of the present invention. As shown in FIG. 16, the speech recognition support system 500 of the present invention includes a storage unit 501 and a speech recognition location suitability information output unit 502.

  The storage unit 501 stores location information indicating a location and information indicating the accuracy of the speech recognition result for the location in association with each other. The storage unit 501 is disclosed as the recognition location matching DB 18 in the above embodiment.

  The voice recognition place suitability information output unit 502 is a place designated as a place to input the speech to be recognized based on the place information stored in association with the storage unit 501 and the information indicating the accuracy of the voice recognition result. Outputs whether or not is suitable for voice recognition.

  Here, “output” includes message transmission in units of processing units within one apparatus and output of data between modules. In addition, when information is passed to other means via the storage means, the case of writing to the storage means is also included.

  In addition, the speech recognition support system 500 may use the noise level included in the speech detected by the speech recognition process for the speech input at a certain location as information indicating the accuracy of the speech recognition result for the location. . For example, the storage unit 501 associates the noise level included in the voice detected by the voice recognition process with respect to the voice input at a certain location with the location information as information indicating the accuracy of the voice recognition result for the location. May be stored.

  In addition, the storage unit 501 stores information indicating the accuracy of the speech recognition result for the location, in association with the location information, and divided into predetermined categories classified based on the date and time information when the speech is input. The recognition place suitability information output unit 502 belongs to a predetermined category in which the current date and time is classified among the information indicating the accuracy of the speech recognition result stored in association with the place information of the place specified in the storage unit 501. Based on the information indicating the accuracy of the speech recognition result, information indicating whether or not the designated location is suitable for speech recognition may be output.

  The speech recognition location suitability information output unit 502 designates information indicating the accuracy of the speech recognition result when information indicating the accuracy of the speech recognition result for the location specified in the storage unit 501 is stored. It may be output as information as to whether or not the determined location is suitable for speech recognition.

  The voice recognition location suitability information output unit 502 is indicated by information indicating the accuracy of the voice recognition result when information indicating the accuracy of the voice recognition result for the location specified in the storage unit 501 is stored. It may be determined whether or not the accuracy of the speech recognition result satisfies a predetermined condition, and the determination result may be output as information as to whether or not the designated location is a location suitable for speech recognition.

  FIG. 17 is a block diagram showing another configuration example of the speech recognition support system 500 according to the present invention. As shown in FIG. 17, the voice recognition support system 500 may further include a voice recognition suitability score calculation unit 503 and a nearest match place presentation unit 504.

  The voice recognition suitability score calculation means 503 is a voice that indicates the degree of suitability of the place indicated by the place information for voice recognition based on the place information indicating the place where the voice is input and the information indicating the accuracy of the voice recognition result for the voice. A recognition fitness score is calculated. The speech recognition suitability score calculation unit 503 is disclosed as the speech recognition unit suitability score calculation unit 14 in the above embodiment.

  In such a case, the speech recognition suitability score calculated by the speech recognition suitability score calculation unit 503 may be used as information indicating the accuracy of the speech recognition result for the place. For example, the storage unit 501 may store the speech recognition suitability score calculated by the speech recognition suitability score calculation unit 503 in association with the location information as information indicating the accuracy of the speech recognition result for the location. Note that the location information associated here does not have to completely match the location information indicating the location where the voice is input. For example, location information that is the same location as the location where the voice is input may be used according to a certain standard.

  Further, for example, the speech recognition suitability score calculation unit 503 regards the amount of correction performed by the user for the speech recognition result of speech spoken by a certain user as information indicating the accuracy of the speech recognition result, and A speech recognition suitability score for the speech input location may be calculated.

  Further, the voice recognition conformity score calculation means 503 uses the difference between the correction amount performed by the user on the voice recognition result of the voice spoken by a certain user and the average of the past correction amount of the user. Then, a speech recognition suitability score for the speech input location may be calculated.

  Further, the speech recognition suitability score calculating unit 503 regards the noise level included in the speech detected by the speech recognition process for the speech spoken by the user as information indicating the accuracy of the speech recognition result of the speech, and A speech recognition suitability score for the speech input location may be calculated.

  In addition, the speech recognition suitability score calculation unit 503 may calculate the speech recognition suitability score by dividing into predetermined categories classified based on the date and time information when the speech is input. The storage unit 501 stores the speech recognition conformity score calculated by the speech recognition conformity score calculation unit 503 in association with the location information and divided into predetermined categories classified based on the date and time information when the speech is input. May be. Then, the voice recognition place suitability information output unit 501 outputs information on whether or not the designated place is suitable for voice recognition based on the voice recognition suitability score belonging to a predetermined category into which the current date and time is classified. May be.

  The nearest matching place presenting means 504 is configured so that the designated place is suitable for voice recognition based on the information on whether or not the designated place output by the voice recognition place suitability information output means 502 is suitable for voice recognition. If it is determined that there is not, information on a place suitable for speech recognition close to the designated place is output. In the above embodiment, the nearest suitable place presentation unit 504 is disclosed as the nearest optimum place presentation unit 16.

  The nearest compatible place presenting means 504 outputs information on a place where the distance from the designated place is within a predetermined range and the voice recognition conformity score for the place satisfies a predetermined condition. May be.

  The present invention is preferably applicable to any system that can perform speech recognition by inputting speech at an unspecified location.

DESCRIPTION OF SYMBOLS 100 Speech recognition support server 11 Speech recognition part 12 Recognition result text display part 13 Recognition result text correction part 14 Speech recognition conformity score calculation part 15 Speech recognition conformity place judgment part 16 Nearest optimum place presentation part 17 Speech recognition result database 18 Speech recognition Applicable location database 19 User database 200 Network 301, 302, 303 User terminal 31 Transmission / reception unit 32 Voice input unit 33 Display unit 34 Location information input unit 500 Voice recognition support system 501 Storage unit 502 Voice recognition location suitability information output unit 503 Audio Recognition conformity score calculation means 504 Nearest conformity place presentation means

Claims (35)

Storage means for associating and storing place information indicating a place and information indicating accuracy of a speech recognition result for the place;
Whether or not a location designated as a location for inputting speech to be recognized is suitable for speech recognition based on location information stored in association with the storage means and information indicating accuracy of speech recognition results A speech recognition support system comprising speech recognition place suitability information output means for outputting information. A voice for calculating a voice recognition suitability score indicating a degree of suitability of the place indicated by the location information for voice recognition based on the place information indicating the place where the voice is input and the information indicating the accuracy of the voice recognition result for the voice. It has a recognition fitness score calculation means,
The speech recognition support system according to claim 1, wherein the speech recognition suitability score calculated by the speech recognition suitability score calculation unit is used as information indicating accuracy of a speech recognition result for the place. The speech recognition conformity score calculating means regards the amount of correction made by the user for the speech recognition result of speech spoken by a certain user as information indicating the accuracy of the speech recognition result for the speech, and inputs the speech The speech recognition support system according to claim 2, wherein a speech recognition conformity score for a place is calculated. The voice recognition conformity score calculation means uses the difference between the correction amount performed by the user on the voice recognition result of the voice spoken by a certain user and the average of the past correction amount of the user, The speech recognition support system according to claim 3, wherein a speech recognition conformity score for the speech input location is calculated. The voice recognition conformity score calculating means regards the noise level included in the voice detected by the voice recognition process for the voice input at a certain place as information indicating the accuracy of the voice recognition result for the voice, The speech recognition support system according to claim 2, wherein a speech recognition conformity score for the input location is calculated. The speech recognition support system according to claim 1, wherein a noise level included in the speech detected by speech recognition processing for speech input at a certain location is used as information indicating accuracy of a speech recognition result for the location. The storage means associates with the location information and stores information indicating the accuracy of the speech recognition result for the location divided into predetermined categories classified based on the date and time information when the speech is input,
The voice recognition place suitability information output means is arranged in the predetermined category into which the current date and time are classified among the information indicating the accuracy of the voice recognition result stored in association with the place information of the place designated in the storage means. The voice recognition according to any one of claims 1 to 6, wherein information indicating whether or not the designated place is suitable for voice recognition is output based on information indicating accuracy of a voice recognition result to which the voice recognition belongs. Support system. The speech recognition location suitability information output means, when information indicating the accuracy of the speech recognition result for the specified location is stored in the storage means, the information indicating the accuracy of the speech recognition result is stored in the specified location The voice recognition support system according to any one of claims 1 to 7, wherein the voice recognition support system outputs the information as to whether the place is suitable for voice recognition. The speech recognition location suitability information output means, when information indicating the accuracy of the speech recognition result for the location specified in the storage means is stored, the speech recognition result of the speech recognition result indicated by the information indicating the accuracy of the speech recognition result The determination is made as to whether or not the accuracy satisfies a predetermined condition, and the result of the determination is output as information as to whether or not the designated location is suitable for speech recognition. The speech recognition support system according to claim 1. Specified when it is determined that the designated location is not suitable for speech recognition based on the information on whether or not the designated location output by the speech recognition location suitability information output means is suitable for speech recognition. The voice recognition support system according to any one of claims 1 to 9, further comprising a nearest matching place presentation unit that outputs information on a place suitable for voice recognition close to a different place. The nearest compatible place presentation means outputs information on a place whose distance from the designated place is within a predetermined range and the accuracy of the voice recognition result for the place satisfies a predetermined condition. Item 11. The speech recognition support system according to Item 10. Storage means for associating and storing place information indicating a place and information indicating accuracy of a speech recognition result for the place;
Whether or not a location designated as a location for inputting speech to be recognized is suitable for speech recognition based on location information stored in association with the storage means and information indicating accuracy of speech recognition results A speech recognition support apparatus comprising speech recognition place suitability information output means for outputting information. A voice for calculating a voice recognition suitability score indicating suitability for voice recognition of the place indicated by the location information based on the location information indicating the place where the voice is input and the information indicating the accuracy of the voice recognition result for the voice. It has a recognition fitness score calculation means,
The speech recognition support device according to claim 12, wherein the speech recognition suitability score calculated by the speech recognition suitability score calculation unit is used as information indicating accuracy of a speech recognition result for the place. The speech recognition support device according to claim 12, wherein a noise level included in the speech detected by speech recognition processing for speech input at a location is used as information indicating accuracy of a speech recognition result for the location. The storage means associates with the location information and stores information indicating the accuracy of the speech recognition result for the location divided into predetermined categories classified based on the date and time information when the speech is input,
The voice recognition place suitability information output means is arranged in the predetermined category into which the current date and time are classified among the information indicating the accuracy of the voice recognition result stored in association with the place information of the place designated in the storage means. The voice recognition according to any one of claims 12 to 14, wherein information indicating whether or not the designated place is suitable for voice recognition is output based on information indicating accuracy of a voice recognition result to which the voice recognition belongs. Support device. The speech recognition location suitability information output means, when information indicating the accuracy of the speech recognition result for the specified location is stored in the storage means, the information indicating the accuracy of the speech recognition result is stored in the specified location The speech recognition support device according to any one of claims 12 to 15, wherein the speech recognition support device outputs the information as to whether the place is suitable for speech recognition. When it is determined that the designated location is not suitable for speech recognition based on the information on whether the designated location is suitable for speech recognition, which is output by the speech recognition location suitability information output unit, The speech recognition support apparatus according to any one of claims 12 to 16, further comprising a nearest matching place presentation unit that outputs information on a place suitable for speech recognition close to a designated place. When location information, which is a location for inputting speech to be recognized, is input, the location information is transmitted to a predetermined server device, and a speech recognition result for the location associated with the location information is transmitted. A location recognition result accuracy information receiving means for receiving information indicating accuracy;
Based on the location information and the information indicating the accuracy of the speech recognition result for the location received by the location recognition result accuracy information receiving means, a location designated as a location for inputting the speech to be recognized is used for speech recognition. A user terminal comprising voice recognition location suitability information output means for outputting information on suitability. Voice recognition result accuracy that collects information indicating the accuracy of the voice recognition result for each voice input at the location indicated by the input location information when the location information that is the location where the speech to be recognized is input is input Information gathering means;
Whether or not a location designated as a location for inputting speech to be recognized is suitable for speech recognition based on information indicating accuracy of speech recognition results for each speech collected by the speech recognition result accuracy information collecting means And a voice recognition location suitability information output means for outputting the information. Speech recognition for calculating a speech recognition conformity score indicating suitability for speech recognition of the location indicated by the input location information based on information indicating the accuracy of the speech recognition result collected by the speech recognition result accuracy information collecting means A fitness score calculation means,
The voice recognition place suitability information output means is based on the voice recognition suitability score calculated by the voice recognition suitability score calculation means to determine whether or not the place designated as the place to input the speech to be recognized is suitable for voice recognition. The user terminal according to claim 19. When it is determined that the designated location is not suitable for speech recognition based on the information on whether the designated location is suitable for speech recognition, which is output by the speech recognition location suitability information output unit, The user terminal according to any one of claims 18 to 20, further comprising a nearest matching place presentation unit that outputs information on a place suitable for speech recognition close to a designated place. The location information input means for inputting the current location information of the user terminal as the location information that is used as a location for inputting the speech to be recognized is described in any one of claims 18 to 21. User terminal. The location information indicating the location and the information indicating the accuracy of the speech recognition result for the location are associated with each other and stored in a predetermined storage device,
Whether or not a location designated as a location for inputting speech to be recognized is suitable for speech recognition based on location information stored in association with the storage device and information indicating accuracy of speech recognition results A speech recognition support method characterized by outputting information. Based on the location information indicating the location where the speech is input and the information indicating the accuracy of the speech recognition result for the speech, a speech recognition suitability score indicating the suitability for speech recognition of the location indicated by the location information is calculated,
The speech recognition support method according to claim 23, wherein the calculated speech recognition conformity score is used as information indicating accuracy of a speech recognition result for the place. Considering the amount of correction performed by the user for the speech recognition result of speech spoken by a certain user as information indicating the accuracy of the speech recognition result for the speech, the speech recognition fitness score for the input location of the speech is The speech recognition support method according to claim 24. Considering the noise level included in the voice detected by the voice recognition process for the voice input at a certain place as information indicating the accuracy of the voice recognition result for the voice, the voice recognition conformity score for the voice input place The speech recognition support method according to claim 24. The speech recognition support method according to claim 23, wherein a noise level included in the speech detected by speech recognition processing for speech input at a location is used as information indicating accuracy of a speech recognition result for the location. Corresponding to the location information, and storing the information indicating the accuracy of the speech recognition result for the location, divided into predetermined categories classified based on the date and time information when the voice is input,
Based on the information indicating the accuracy of the speech recognition result belonging to the predetermined category in which the current date and time is classified among the information indicating the accuracy of the speech recognition result stored in association with the location information of the designated location, The speech recognition support method according to any one of claims 23 to 27, wherein information indicating whether or not the designated location is suitable for speech recognition is output. If the information indicating the accuracy of the speech recognition result for the designated location is stored in the predetermined storage device, the information indicating the accuracy of the speech recognition result is determined based on whether the designated location is suitable for speech recognition. The speech recognition support method according to any one of claims 23 to 28, which is output as information on whether or not. When information indicating the accuracy of the speech recognition result for the designated location is stored in the predetermined storage device, the accuracy of the speech recognition result indicated by the information indicating the accuracy of the speech recognition result satisfies a predetermined condition. The voice according to any one of claims 23 to 28, wherein whether or not the designated place is a place suitable for voice recognition is output. Recognition support method. Suitable for speech recognition close to the specified location when it is determined that the specified location is not suitable for speech recognition based on the output information on whether the specified location is suitable for speech recognition The speech recognition support method according to any one of claims 23 to 30, wherein information on the location of the location is output. On the computer,
Location information indicating the location and information indicating the accuracy of the speech recognition result for the location are stored in a predetermined storage device in association with each other, and location information used as a location for inputting the speech to be recognized is input. A speech recognition support program for executing a location recognition result accuracy information output process for outputting information indicating the accuracy of the speech recognition result for the designated location stored in the storage device. On the computer,
A storage process in which location information indicating a location and information indicating the accuracy of a voice recognition result for the location are associated with each other and stored in a predetermined storage device, and the location information and audio stored in association with the storage device A voice that executes voice recognition location suitability information output processing that outputs information indicating whether or not a location designated as a location to which speech to be recognized is input is suitable for voice recognition based on information indicating accuracy of the recognition result Recognition support program. On the computer,
When location information, which is a location for inputting speech to be recognized, is input, the location information is transmitted to a predetermined server device, and a speech recognition result for the location associated with the location information is transmitted. A reception process for receiving information indicating accuracy, and a location designated as a location for inputting recognition target speech based on the location information and information indicating accuracy of a speech recognition result for the received location. A speech recognition support program for executing speech recognition place suitability information output processing for outputting information on whether or not it is suitable for speech recognition. On the computer,
A collection process for collecting information indicating accuracy of a speech recognition result for a voice input at a location indicated by the input location information when location information that is a location for inputting a speech to be recognized is input; and Based on the location information and the information indicating the accuracy of the collected speech recognition result, speech recognition that outputs information indicating whether or not the location designated as the location for inputting the speech to be recognized is suitable for speech recognition A speech recognition support program that executes location adequacy information output processing.

Images