JP2004226910A - Device, method, and program for speech recognition error correction - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To make efficient operation for correcting a speech recognition error of text data as a speech recognition result without taking into consideration the balance between the numbers of finders and correctors. <P>SOLUTION: A speech recognition error correcting device comprises a character data reception part 15, a correction information reception part 17, a correcting terminal operation decision part 19, a character presenting speed variation part 21, a character attribute change part 23, a character attribute information transmission part 25, a display area setting part 27, an unnecessary character insertion part 29, a presentation information transmission part 31, a text data protection part 33, a character string integration part 35, and a screen display information transmission part 39. A speech reproduction part 3 comprises a speech reception part 5, a speech storage part 7, a speech data transmission part 9, and a speech presenting speed variation part 11. Such a speech recognition error correcting device 1 enables a corrector to make corrections alone without partially allotting the correcting operation to the finder and a corrector of a speech recognition error. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a speech recognition error correction device for correcting a speech recognition error, a speech recognition error correction method, and a speech recognition error correction program.
[0002]
[Prior art]
2. Description of the Related Art Conventionally, a speech recognition error correction device using a speech recognition technology includes a voice captioning device and a character data correction device (see Patent Documents 1 and 2).
In the former, when correcting a speech recognition error in order to convert the speech recognition result into subtitles in real time, the timing between the presentation of the speaker's speech used for speech recognition and the presentation of text data as the speech recognition result In this case, the efficiency of the operation of correcting the speech recognition error is improved by optimizing the speech recognition.
The latter is a pointing device (pointing terminal) that finds and selects a speech recognition error in text data (speech recognition result) output from a text data output device (speech recognition device) in order to convert the speech recognition result into subtitles in real time. The device and the correction terminal (correction terminal device) that corrects and replaces the error in the speech recognition are assigned roles, thereby improving the efficiency of the operation of correcting the error in the speech recognition result.
[0003]
[Patent Document 1]
JP 2001-142482 A (paragraphs 0018 to 0043,
(Fig. 1)
[Patent Document 2]
JP 2001-60192 A (paragraph numbers 0023 to 0043,
(Fig. 1)
[0004]
[Problems to be solved by the invention]
However, in the conventional speech recognition error correction device, the role of work is shared between a discovery operator (hereinafter, referred to as a discoverer) for finding an error in speech recognition and a correction operator (hereinafter, referred to as a corrector) for correcting the error. As a result, the error in the speech recognition result is corrected, so that the following disadvantages arise from the difference in the work contents of the discoverer and the corrector.
[0005]
1. Consideration must be given to the balance between the number of discoverers and the number of correctors, causing inconvenience in personnel management such as staffing.
2. Need to provide separate training for discoverers and correctors.
3. When using the speech recognition error correction device to correct errors in speech recognition results, the discoverer and the corrector must always work in pairs, so they must work simultaneously and in the same place. Not work efficiency.
In addition, there is a problem that the work place is restricted.
[0006]
Therefore, the present invention solves such a problem without considering the balance between the number of discoverers and correctors, and without the need for separate education and training and minimizing the restrictions on the work place. It is an object of the present invention to provide a speech recognition error correction device, a method thereof, and a program capable of performing an error correction operation efficiently.
[0007]
[Means for Solving the Problems]
The present invention has the following configuration to achieve the above object.
The invention according to claim 1 receives the speech data output from the speech recognition device as the speech recognition target and the text data as the speech recognition result, and detects a speech recognition error included in the text data. A speech recognition error correction device for correcting by a plurality of terminals, a presentation unit for presenting the voice data and the text data to the terminal, and a text data presented to the terminal by the presentation unit. Data receiving means for receiving the indicated data indicating the corrected character range indicated and represented by the corrected text data obtained by correcting the text data of the corrected character range, and based on the indicated data received by the data receiving means, The correction sentence until the correction by the pointing data transmission terminal which is the terminal that transmitted the pointing data is completed. Text data protection means for protecting the range from indications by a terminal other than the indication data transmission terminal; and output means for correcting the text data based on the corrected text data received by the data reception means and outputting the text data to each of the terminals. And the following.
[0008]
According to this configuration, the voice data and the text data output from the voice recognition device are received and presented to the plurality of correction terminals. Then, from these terminals, the indication data indicating the corrected character range indicating the speech recognition error and the corrected text data corrected are received, and the text data protection unit transmits the indicated data to the terminal other than the indicated data transmitting terminal that transmitted the indicated data. Until the correction by the data transmitting terminal is completed, a speech recognition error cannot be pointed out. In other words, the pointing data transmitting terminal that has pointed out the voice recognition error first corrects the voice recognition error with the highest priority, receives the corrected text data transmitted from the pointing data transmitting terminal, and corrects the text data. . Then, the latest correction information is sequentially transmitted to each terminal and notified by the output means, and the correction work is continued.
If the corrector using the terminal incorrectly points out the text data that is the correct speech recognition result as the corrected character range, the correct character range may be selected by, for example, selecting a delete key. Can be returned to the state before the operation.
[0009]
The invention according to claim 2 is the speech recognition error correction device according to claim 1 or 2, wherein the plurality of terminals are connected to the speech recognition error correction device via a network. I do.
[0010]
According to this configuration, the speech recognition error correction device is operated by each terminal connected via the network. Therefore, it is possible to flexibly cope with the setting place of each terminal, that is, the setting of the correction work place.
[0011]
According to a third aspect of the present invention, in the speech recognition error correction device according to the first or second aspect, the presenting unit includes a display color, a display size, a character By changing the attribute of a character including at least one of the types, a pointing character attribute changing function for presenting the change of the character attribute to a terminal other than the pointing data transmitting terminal; When the data is being modified, a modified character attribute change function that sets a character attribute of the indicated data being modified to a character attribute different from the character attribute changed by the indicated character attribute change function, I do.
[0012]
According to this configuration, the attribute of the character indicating the speech recognition error portion in the text data as the speech recognition result is changed by the indicated character attribute changing function, and the terminal other than the indicated data transmission terminal indicating the speech recognition error is changed. , The text data of the changed character attribute is presented. Then, the corrected character attribute change function changes the attribute of the character being corrected to a character attribute other than the character attribute changed at the time of pointing out the character of speech recognition error, However, the text data of the changed character attribute is presented.
[0013]
According to a fourth aspect of the present invention, in the speech recognition error correction device according to the first or second aspect, the presenting means is configured to display color of at least particles of a part of speech related to the text data as the speech recognition result. , A character part of speech attribute changing function for changing an attribute of a character including at least one of a display size and a character type.
[0014]
According to such a configuration, the character attribute change function changes the character attribute of at least the particles of the part of speech of the text data as the speech recognition result. Arousing can be encouraged.
[0015]
According to a fifth aspect of the present invention, in the speech recognition error correction device according to the first or second aspect, the presenting unit includes a character string relating to text data as the speech recognition result presented to the terminal. A display area setting function for arbitrarily setting at least one of the display width of the display area and the background color.
[0016]
According to this configuration, the display area setting function sets (changes) at least one of the display width and the background color in the display area of the text data as the speech recognition result based on a request from the user of the terminal. If this is the case, the display area of the text data becomes easier to see, and the viewability of the display screen is improved.
[0017]
According to a sixth aspect of the present invention, in the speech recognition error correction device according to the first or second aspect, the presenting unit includes an unnecessary character in text data as the speech recognition result presented to the terminal. It is characterized by having an unnecessary character insertion function for inserting a.
[0018]
According to this configuration, the unnecessary character insertion function inserts unnecessary characters into the text data that is the result of speech recognition, so that the corrector using the terminal needs to delete the unnecessary characters, Even when the speech recognition error rate of the presented text data is low, the text data to be corrected is small, and the monotonous work continues, the concentration of the correction work can be maintained.
[0019]
The invention according to claim 7 is configured to receive the speech data output from the speech recognition device as the speech recognition target and the text data as the speech recognition result, and to recognize the speech recognition error included in the text data. A method for correcting a speech recognition error by a plurality of terminals, the method comprising: presenting the voice data and the text data to the terminal; and providing the terminal with text data presented to the terminal by the presenting step. A data reception step of receiving corrected text data obtained by correcting the text data of the corrected character range indicated by the corrected character range indicated and represented by, and based on the specified data received in the data receiving step, Correction by the pointing data transmission terminal, which is the terminal that transmitted the pointing data, is completed. Up to the modified character range, a text data protection step to protect from indications by a terminal other than the indication data transmission terminal, and the text data is modified based on the modified text data received in the data receiving step, and each of the terminals And an output step of outputting.
[0020]
According to the speech recognition error correction method, first, the speech data output from the speech recognition device and the text data as the speech recognition result in the presentation step are presented to the terminal. Subsequently, in the data receiving step, of the text data presented to the terminal by the presentation step, the pointing data indicating the corrected character range in which the user of the terminal has pointed out a speech recognition error and the corrected text data corrected are received. . Then, in the text data protection step, based on the pointed data, the text data of the speech recognition error is protected from the pointed out by a terminal other than the pointed out data transmitting terminal until the correction work by the pointed out data transmitting terminal is completed. I do. Then, the text data relating to the speech recognition error is corrected based on the corrected text data received in the data receiving step. In other words, the pointing data transmitting terminal that has pointed out the voice recognition error first corrects the voice recognition error with the highest priority, receives the corrected text data transmitted from the pointing data transmitting terminal, and corrects the text data. . Next, in the output step, the corrected text data is sequentially transmitted to each terminal as the latest correction information, and the correction work is continued.
[0021]
The invention according to claim 8 is configured to receive the speech data output from the speech recognition device and the text data as the speech recognition result, and to recognize the speech recognition error included in the text data. A speech recognition error correction device to be corrected by a plurality of terminals, presentation means for presenting the speech data and the text data to the terminal, and text data presented to the terminal by the presentation means, indicated by the terminal. Data receiving means for receiving indication data indicating a corrected character range represented by text data and corrected text data obtained by correcting text data in the corrected character range, and transmitting the indication data based on the indication data received by the data receiving means. Until the correction by the indicated data transmission terminal is completed, the corrected character range, The text data protecting means protects from the indication by a terminal other than the indication data transmitting terminal, and the output means for correcting the text data based on the corrected text data received by the data receiving means and outputting the corrected text data to each of the terminals. It is characterized by the following.
[0022]
According to such a speech recognition error correction program, a function as a speech recognition error correction device can be generated and executed according to the processing procedure of the program, so that the speech data and text data output from the speech recognition device can be received. The modified data is presented to the plurality of correction terminals, and the user of these terminals receives the indicated data indicating the corrected character range indicating the voice recognition error and the corrected data. On the basis of this, it is impossible to point out a speech recognition error for a terminal other than the terminal that transmitted the indicated data until the correction work by the indicated data transmitting terminal is completed. As a result, the work of correcting the speech recognition error is performed with the highest priority by the indication data transmitting terminal that has indicated the speech recognition error first. Then, the text data is corrected based on the received corrected text data transmitted from the indication data transmitting terminal, and the corrected text data is sequentially transmitted to each terminal as the latest correction information, and the correction work is continued.
[0023]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
(Outline of speech recognition error correction system)
First, an outline of a speech recognition error correction system will be described with reference to FIG.
FIG. 1 is a schematic diagram showing a configuration of a speech recognition error correction system according to an embodiment of the present invention.
[0024]
The speech recognition error correction system according to one embodiment of the present invention converts text data, which is the result of speech recognition of a speech uttered by a speaker by the speech recognition device 2, and speech data targeted for speech recognition via a network. And a speech recognition error correcting device 1 that integrates the corrected text data corrected by these correcting terminals 41, and points out a corrected character range from the text data that is the voice recognition result. And a correction terminal 41 for correcting the characters to correct characters.
In the present embodiment, the correction terminal 41 corresponds to the terminal described in the claims.
[0025]
Further, in the present embodiment, the speech recognition device 2 is used as an example of a device that outputs text data to be corrected by the speech recognition error correction device 1, but any device that outputs text data can use a word processor function or the like. A personal computer or the like equipped with a voice recognition function may be used.
[0026]
In the present embodiment, the text data is corrected to corrected text data by the correction terminals 41 (41A, 41B, 41C) connected to the speech recognition error correction device 1, and the corrected text data is corrected. Receives and corrects the text data on the speech recognition error correction device 1, and outputs the corrected integrated text data in real time (this is the operation of the speech recognition error correction system).
Then, the corrected integrated text data is transmitted as a text broadcast to the television 45 or a personal computer equipped with the function as the television 45 via the broadcast communication network (broadcast wave EW) using the subtitle transmission device 43. The case of broadcasting will be described as an example.
[0027]
Hereinafter, in this embodiment, the pointing data (details described later) indicating the speech recognition error included in the text data transmitted from the speech recognition error correction device 1 is transmitted to the speech recognition error correction device 1. The correction terminal 41 will be referred to as a pointing data transmission terminal 41.
[0028]
(Configuration of speech recognition error correction device)
Next, the configuration of the speech recognition error correction device 1 will be described with reference to FIGS.
FIG. 2 is a block diagram showing the configuration of the speech recognition error correction device 1 according to one embodiment of the present invention.
As shown in FIG. 1, the voice recognition error correction device 1 includes a text data correction unit 13 and a voice reproduction unit 3, and is connected to correction terminals 41 (41A, 41B, 41C) via a network.
[0029]
As shown in FIG. 2, the text data correction unit 13 includes a character data reception unit 15, a correction information reception unit 17, a correction terminal operation determination unit 19, a character presentation speed variable unit 21, a character attribute change unit 23, and character attribute information transmission. It comprises a unit 25, a display area setting unit 27, an unnecessary character insertion unit 29, a presentation information transmission unit 31, a text data protection unit 33, a character string integration unit 35, a screen display information transmission unit 37, and a subtitle output configuration unit 39. .
[0030]
In this embodiment, the character data receiving unit 15 and the correction information receiving unit 17 correspond to the data receiving unit described in the claims, and the character presentation speed varying unit 21 and the character attribute changing unit 23 correspond to the claims. The text data protection unit 33 corresponds to the text data protection unit described in the claims, and the character attribute information transmission unit 25 and the presentation information transmission unit 31 correspond to the claims in the claims. The display area setting section 27 corresponds to the display area setting function described in the claims, and the unnecessary character insertion section 29 corresponds to the unnecessary character insertion function described in the claims. is there.
[0031]
The character data receiving unit 15 receives text data as a result of speech recognition by the speech recognition device 2 (see FIG. 1).
The correction information receiving unit 17 outputs the corrected text data obtained by correcting the corrected text ID (Identification), the corrected terminal ID, the control information, and the speech recognition error of the speech recognition device 2 to the correct text data by the correction terminal 41 from the correction terminal 15. To receive.
[0032]
Note that the “corrected text ID” is an identification symbol (pointed data, which will be described later) for identifying a corrected character range that points out an error portion of text data that is a speech recognition result when the correction terminal 41 enters a correction operation. That is. The “correction terminal ID” is an identification symbol for identifying the correction terminal 41 that has pointed out a correction character range that is an error part of text data that is a speech recognition result when the correction terminal 15 enters a correction operation. . The “control information” is information received from the correction terminal 41 by the correction information receiving unit 17, information on the presentation operation of the text data to the correction terminal 41, information serving as a reference of a control command relating to the reproduction speed or stop, and the correction terminal 41. This is information that serves as a reference for a control instruction regarding the reproduction speed or stop of the audio data presentation operation.
[0033]
The correction terminal operation determination unit 19 has a function of transmitting the correction terminal ID and the control information received from the correction information reception unit 17 to the character presentation speed variable unit 21 (details will be described later), and presents voice data to the correction terminal 41. It has a function of transmitting character presentation timing information and control information serving as timing references to the voice presentation speed variable unit 11 (details will be described later). Further, when the presentation of the text data or the voice data to the correction terminal 41 is temporarily stopped, a delay in the pause is determined, and a high-speed presentation instruction of the text data or the voice data to the correction terminal 41 is performed. . Further, when all of the correction terminals 41 start the correction work, a command for temporarily stopping the presentation of the text data and the voice data to the correction terminal 41 is issued to the character presentation speed variable unit 21 (described later) and the voice presentation speed variable unit 11 (described later). To).
[0034]
Note that the “character presentation timing information” is reference information that is a timing for presenting voice data to the correction terminal 41 for supporting the operation of correcting text data in the correction character range to correct characters. The voice presentation speed variable unit 11 (described later) controls the timing of the presentation of the voice data to the correction terminal 41 based on the character presentation timing information.
[0035]
The character presentation speed variable unit 21 corrects a control command related to the reproduction speed or stop of the presentation operation of the text data to the correction terminal 41 based on the correction terminal ID and the control information received from the correction terminal operation determination unit 19. The ID is transmitted to the correction terminal 41 related to the ID. Further, when all the correction terminals 41 have started the correction work, an instruction to suspend the presentation of the text data to the correction terminal 41 is received from the correction terminal operation determination unit 19, and all the correction terminals 41 (41A, 41B) are received. , 41C).
[0036]
The character attribute changing unit 23 includes a pointed character attribute changing unit 23a, a corrected character attribute changing unit 23b, and a character part of speech attribute changing unit 23c.
In the present embodiment, the indicated character attribute changing unit 23a corresponds to the indicated character attribute changing function described in the claims, and the modified character attribute changing unit 23b corresponds to the modified character attribute changing function described in the claims. And the character part-of-speech attribute changing unit 23c corresponds to a character part-of-speech attribute changing function described in the claims.
[0037]
When the correction terminal 41 points out a correction character range that is text data indicating a speech recognition error by the correction terminal 41, the correction character range according to the correction text ID transmitted from the correction terminal 41 to the correction information receiving unit 17 With respect to the character (1), the character attribute including at least one of a display color, a display size, and a character type is changed for the corrected text data received by the correction information receiving unit 17 from the correction terminal 41. Also, by presenting the corrected text data in which the character attribute of the corrected character range has been changed to the correction terminals 41 other than the correction terminal 41 (pointed data transmission terminal 41) that transmitted the indicated data (corrected text ID), It is also possible to check the indication status related to the indication data from a modification terminal other than the modification terminal 41 (the indication data transmission terminal 41).
[0038]
The “pointed data” refers to a corrected character range pointed out by the pointed data transmitting terminal 41 or data indicating the pointed data transmitting terminal 41, and specifically, the corrected text ID or the corrected terminal ID described above. The pointed-out data transmitting terminal 41 that has transmitted the corrected text ID transmits the corrected terminal ID to the correction information receiving unit 17.
In addition, for changing the character attribute, for example, the font is changed to italic, bold, shadowing, three-dimensional character, bag character, etc. , Underlining, character enclosing, and the like.
[0039]
The corrected character attribute changing unit 23b changes the character whose correction character range is being corrected to a character attribute other than the character attribute whose character attribute of the correction character range has been changed by the indicated character attribute changing unit 23a. In addition, the character attributes of the characters being corrected in the corrected character range are changed to the correction terminals 41 other than the terminal 41 whose correction character range is being corrected (the pointing data transmission terminal 41 whose text data in the corrected character range is being corrected). By presenting the corrected text data, it is possible to check the correction status related to the indicated data from the correction terminals 41 other than the terminal 41 that is correcting.
Note that the change of the character attribute is the same as the description in the pointed-out character attribute changing unit 23a, and therefore, the description thereof is omitted.
[0040]
The character part-of-speech attribute changing unit 23c, as shown in FIG. 4, displays the character display color, the display size, and the character size of at least particles of the part of speech of the character string in the text data transmitted from the character data The character attribute including at least one of the types is changed to change the corrected text data received by the correction information receiving unit 17 from the correction terminal 41 (pointing data transmission terminal 41). In particular, particles have a high frequency of occurrence, and it is very common to overlook the corrector's discovery of speech recognition errors. Is what you do. Changing the character attribute to enlarge the character display of particles is particularly effective when using a touch panel as a pointing device because it increases the accuracy of finding particle recognition errors in particles and makes it easier to point out (select) on the touch panel. It is.
[0041]
It should be noted that a control unit (not shown) may be initially set to transmit text data to the correction terminal 41 after changing the character attribute of at least particles of the part of speech of the character string in the text data.
Note that the change of the character attribute is the same as the description in the pointed-out character attribute changing unit 23a, and therefore, the description thereof is omitted.
[0042]
Returning to FIG. 2, the description will be continued.
The character attribute information transmitting unit 25 sends the character attribute information, which is the information of the character attribute changed by the indicated character attribute changing unit 23a, the corrected character attribute changing unit 23b, and the character part of speech attribute changing unit 23c, to each correction terminal 41. What to send.
[0043]
As shown in FIG. 5A, the display area setting unit 27 displays the display area AR of the text data, which is the speech recognition result presented to the correction terminal 41, in order to improve the visibility of the display screen. At least one of the width and the background color can be arbitrarily set.
[0044]
Returning to FIG. 2, the description will be continued.
The unnecessary character insertion unit 29 inserts unnecessary characters into text data that is the result of speech recognition presented to the correction terminal 41. FIG. 5B shows an example in which an asterisk mark “*” is inserted into text data as an unnecessary character.
When the speech recognition rate is high and the speech recognition error is small, the correction work is a monotonous work for finding the speech recognition error. As a result, the corrector's concentration on the correction work is reduced, and the accuracy of the correction work is reduced. Therefore, in order to prevent this, unnecessary characters are inserted into the text data as the speech recognition result, and the corrector is allowed to delete the unnecessary characters, thereby maintaining the concentration.
[0045]
Returning to FIG. 2, the description will be continued.
The presentation information transmission unit 31 transmits the display area setting information by the display area setting unit 27 and the unnecessary character insertion information by the unnecessary character insertion unit 29 to the correction terminal 41.
The text data protection unit 33 includes a modified guard terminal determination unit 33a and a modified guard text determination unit 33b.
[0046]
The text data protection unit 33 converts the text data of the speech recognition error based on the pointed data (corrected text ID, corrected terminal ID) received by the correction information receiving unit 17 into a text data other than the pointed data transmission terminal 41 that transmitted the pointed data. From the point indicated by the correction terminal 41 to the completion of the correction work by the pointed data transmission terminal 41.
[0047]
An example of how text data protection is achieved by the text data protection unit 33 will be described with reference to FIG. In FIG. 3, text data of the “frequent terrorist incident at that time”, which is text data of a speech recognition result, is presented to the correction terminal 41A, the correction terminal 41B, and the correction terminal 41C. In this case, the portion of the text data “at that time” (corrected character range) corresponds to a speech recognition error. When the corrector of the correction terminal 41A finds and points out the voice recognition error first, the other correction terminal is used. The corrector 41, that is, the corrector of the correction terminal 41B and the corrector of the correction terminal 41C are guarded so as to be unable to point out the same portion ("at the time") as the portion of the correction character range "at the time". You.
[0048]
At this time, the corrector of the correction terminal 41B and the corrector of the correction terminal 41C can point out portions other than the same portion. Also, even if the corrector of the correction terminal 41A finishes the correction work even in the same place, the corrector of the correction terminal 41B and the corrector of the correction terminal 41C are in the same place (“at that time”). ) And corrective action can be taken.
[0049]
Next, returning to FIG. 2, the components of the text data protection unit 33 will be described.
The correction guard terminal determination unit 33a determines the correction terminal ID sent from the correction information receiving unit 17 (correction pointed out a correction character range that is an error part of text data that is a speech recognition result when the correction terminal 41 starts correction work). Receiving the modified terminal 41 (the modified terminal 41A) indicating the modified character range ("at the time"), and using the identified information as modified guard terminal information. This is transmitted to the terminal 41 (the correction terminal 41B and the correction terminal 41C).
[0050]
The correction guard text determination unit 33b identifies the correction text ID sent from the correction information receiving unit 17 (a correction character range that indicates an error portion of text data that is a speech recognition result when the correction terminal 15 enters a correction operation). ), The modified character range ("at the time") pointed out by the modifying terminal 41 (the modifying terminal 41A) is identified, and the identified information is used as the modified guard text information to modify the modified terminal 41 (the modified terminal 41A). This is transmitted to the terminal 41B and the correction terminal 41C).
[0051]
Next, the character string integrating unit 35 will be described.
The character string integrating unit 35 integrates the corrected text data of each correction terminal 41 received by the correction information receiving unit 17 into one unified sentence based on the corrected text ID. The corrected text integrated data integrated by the character string integrating unit 35 is sequentially fed back to each correction terminal 41, and the latest correction information (corrected integrated text data) is always reflected. The modified integrated text data is used as caption broadcast data by the caption transmission device 43 (see FIG. 1).
[0052]
The screen display information transmitting unit 37 sends the corrected integrated text data in which the corrected text data is integrated into one sentence based on the corrected text ID by the character string integrating unit 35 to each of the correction terminals 41 (41A, 41B, 41C). What to send.
[0053]
The subtitle output forming unit 39 converts the corrected integrated text data into subtitles in order to use the corrected integrated text data in which the corrected text data is integrated into one sentence based on the corrected text ID by the character string integrating unit 35 as a text broadcast. The output data is transmitted to the subtitle transmitting device 43 (see FIG. 1).
[0054]
Next, the audio reproducing unit 3 will be described.
The audio reproduction unit 3 includes an audio reception unit 5, an audio storage unit 7, an audio data transmission unit 9, and an audio presentation speed variable unit 11.
The voice reproduction unit 3 reproduces voice data that has been subjected to voice recognition.
The voice data is transmitted to the correction terminal 41 via the network in synchronization with the character presentation timing information, and the voice output from the headphones connected to the correction terminal 41 and the text data as a voice recognition result are corrected. The corrector of the terminal 41 (41A, 41B, 41C) performs the comparison and collation, thereby supporting the corrector's correction work.
[0055]
The voice receiving unit 5 receives voice data targeted for voice recognition from the voice recognition device 2 (see FIG. 1).
The voice accumulating unit 7 temporarily or for long-term storage for presenting the voice data targeted for voice recognition received by the voice receiving unit 5 to each correction terminal 41.
Generally, a main storage device (main memory) using a semiconductor memory is used for temporary storage, and an external storage device (auxiliary storage device) such as a hard disk, a flexible disk, or a DAT (Digital Audio Tape Recorder) is used for long-term storage. Is used.
[0056]
The audio data transmission unit 9 receives the audio data from the audio storage unit 7 and transmits the audio data to each correction terminal 41 for presentation to each correction terminal 41.
[0057]
The voice presentation speed variable unit 11 is configured to provide the character presentation timing information received from the corrected terminal operation determination unit 19 regarding the operation of presenting the voice data subjected to voice recognition by the voice recognition device 2 (see FIG. 1) to the correction terminal 41. And a control command relating to the reproduction speed or stop based on the control information and the control information. Further, when all the correction terminals 41 start the correction work, the temporary stop instruction of the presentation of the voice data to the correction terminal 41 is received from the correction terminal operation determination unit 19, and all the correction terminals 41 (41A, 41B) are received. , 41C) via the audio data transmission unit 9.
[0058]
Therefore, the voice presentation speed variable unit 11 combines the voice data targeted for voice recognition by the voice recognition device 2 (see FIG. 1) simultaneously with the text data that is the voice recognition result of the voice recognition device 2 (see FIG. 1) ( The text data can be presented to each of the correction terminals 41 (in synchronization with the presentation timing to the correction terminals 41 (41A, 41B, 41C)), and the voice to the correction terminal 41 can be synchronized with the character presentation timing information. Regarding the presentation of the data, it is also possible to pause the presentation and delay the presentation, to perform a high-speed presentation for recovering the delay caused by the suspension, and to perform a replay operation of repeatedly presenting the audio data.
[0059]
Subsequently, returning to FIG. 1, the correction terminal 41 will be described.
The correction terminal 41 (41A, 41B, 41C) is a device for using the voice recognition error correction device 1, and includes a display, a keyboard, a touch panel for designating a range of corrected characters, and headphones for outputting voice data. And a foot input interface (foot pedal) for transmitting a switching signal for speeding up or suspending or resuming the sound data to a control unit (not shown) to control the sound data reproducing operation. Generally, a personal computer is used.
[0060]
The foot pedal includes a foot switch for controlling the stop of reproduction of audio data and an interface such as USB (Universal Serial Bus) or RS232C, and is connected to a personal computer via these interfaces. This foot pedal is used to switch the switch ON / OFF by an operation of "stepping on" or "releasing" of a foot switch (pedal), and to reproduce the contents of the recording tape (audio data) recorded on the recording tape. It is used when controlling the operation.
[0061]
Further, the speech recognition error correction device 1 is connected to a LAN (Local Area Network) or a WAN (Wide Area Network), and can perform any system design irrespective of the installation location of the correction terminal 41. Therefore, a speech recognition error correction operation between remote locations is also possible.
[0062]
As described above, the network may be in any form such as a LAN or WAN, but may be in any form such as a network cable, wireless, or infrared.
However, it is preferable to use a network cable from the viewpoint of safety such as communication packet leakage and high speed processing.
Furthermore, a plurality of correctors (correction terminals 41) can be set by the speech recognition error correction device 1, and the correction character range can be pointed out and corrected by each corrector.
[0063]
(Operation of the speech recognition error correction system)
Next, an example of the operation of the speech recognition error correction system will be described with reference to the sequential chart of FIG. 6 (see FIG. 2 as appropriate).
First, the speech recognition device 2 (see FIG. 1) performs speech recognition on the speech generated by the utterance of the speaker, and transmits text data, which is the result of the speech recognition, to the text data correction unit 13 to determine the speech recognition target. The converted audio data is transmitted to the audio reproducing unit 3 (A1).
[0064]
Here, the speaker is "Hello, this is lunch of news." Said Say, for this speech, voice text data of "Hello, this is news of your Hill" as the speech recognition result of the voice recognition device 2 wrong The case of transmitting to the recognition error correcting device 1 (the text data correcting unit 13 and the voice reproducing unit 3) will be described as an example.
[0065]
Then, text data correction unit 13 of the speech recognition error correction apparatus 1 which has received the voice data, the text data "Hello, news is of your Hill" modify the terminal 41A, and transmits it to the correction terminal 41B and the correction terminal 41C At the same time, it sends the character presentation timing information and control information to the audio reproduction unit 3 (B1). Audio playback unit 3, which has received this, the text data is the result of the voice data "Hello, noon news," which became the object of speech recognition based on the on the character presentation timing information and control information (for example, voice recognition (Simultaneously with presentation to the correction terminal 41A, the correction terminal 41B, and the correction terminal 41C) and the correction terminal 41A, the correction terminal 41B, and the correction terminal 41C (B2).
[0066]
The text data "Hello, your Hill is of news" and the voice data "Hello, noon news" each modification has received the terminal 41 (41A, 41B, 41C) is, by the corrector of correction terminal 41, speech recognition By touching the touch panel, the character "mihiru" (corrected character range) of the error part of "" is pointed out, and the correct character "noon" is corrected on the keyboard.
[0067]
At this time, for example, when the corrector of the correction terminal 41A points out the character "mihiru" of the speech recognition error portion earliest (C1), the pointed-out data (correction text ID, correction terminal ID) is sent from the correction terminal 41A to The text data correcting unit 13 is transmitted to the data correcting unit 13 (C2), and upon receiving this, transmits the corrected guard terminal information and the corrected guard text information to the correcting terminals 41B and 41C (B3). The correction terminal 41B and the correction terminal 41C which have received this cannot perform the correction work during the correction work of the correction terminal 41A, and only the correction terminal 41A can perform the correction work with priority.
[0068]
It should be noted that, when the corrector of correction terminal 41A carries out the modification work, the audio data "Hello, lunch of news" that has been sent from the audio playback unit 3 Modify the person is listening with headphones, are presented in the editing terminals 15A text data "Hello, your Hill is of news" that are increasing the correction accuracy by comparing against the, modification work is carried out.
[0069]
Then, the correction terminal 41A transmits the corrected text data “lunch” as the correction result to the text data correction unit 13 (C3), and the text data correction unit 13 that has received the correction text data converts the corrected text data “lunch” into one. modify integration was integrated into the sentence text data "Hello, lunch of news" to create the, the amendment integrated text data is sequentially, the editing terminals 41A, is transmitted to the correction terminal 41B and the correction terminal 41C latest fixes information is all terminals (As an example of the present embodiment, the correction is performed on the correction terminal 41A, the correction terminal 41B, and the correction terminal 41C), and a correction operation is performed (B4).
[0070]
Next, the presence / absence of the corrected text data is determined (B5), and if there is the corrected text data (B5, YES), the corrected text data is integrated into the corrected integrated text data to make the corrected integrated text data correct. (B6). Then, text data such amendment integrated text data "Hello, this is lunch of news" from the correction unit 13 to the subtitle delivery device 43 is transmitted (B7).
[0071]
When there is no corrected text data (B5, NO), for example, it is determined whether or not one minute has elapsed after transmitting the text data to each correction terminal 41 (B8). If (YES), the process proceeds to B6, and the corrected integrated text data is determined. If the corrected integrated text data has not passed (B8, NO), the process returns to B5.
Although not shown in this flowchart, the steps from A1 to B7 are repeated as a speech recognition error correction system while text data as a speech recognition result is presented. The operation of the error correction system ends.
[0072]
Further, as an example of the present embodiment, the corrected integrated text data transmitted from the caption transmitting device 43 that has received the corrected integrated text data transmitted in B7 to the television 45 via the broadcast wave EW (see FIG. 1). (D1), the television 45, "Hello, noon news" is received as teletext with subtitles (E1) shown in an example of the flowchart of case (inner portion from the broken line shown in the lower left of FIG. 6) Keep it.
[0073]
In the determination of the corrected integrated text data in B6, the text data correcting unit 13 integrates the corrected text data after receiving the corrected text data from the correction terminal 41 that has transmitted the corrected text data at the latest time. It can also be determined when the data is created.
[0074]
The embodiments of the present invention have been described above, but the present invention is not limited to the above-described embodiments, but can be implemented in various forms.
The present invention also relates to a voice recognition error correction method, a voice recognition error correction program for realizing such a voice recognition error correction method in the voice recognition error correction device 1, and a recording medium on which the voice recognition error correction program is recorded. Is what you do.
[0075]
【The invention's effect】
As described above, the present invention has the following effects.
According to the first, seventh and eighth aspects of the present invention, the pointing data indicating the corrected character range indicating the voice recognition error and the corrected text data are received from the terminal connected to the voice recognition error correction device. For text data protection means, for terminals other than the pointed-data transmitting terminal that transmitted the pointed-data, point out a speech recognition error and perform correction work until the correction by the pointed-data transmitting terminal is completed. Can not be.
[0076]
Therefore, the correction work can be performed by the corrector alone without sharing the correction work between the discoverer and the corrector of the corrected character range, which is the speech recognition error part. Therefore, the efficiency of the correction work can be improved. In addition, personnel and labor management such as staffing and securing of personnel is facilitated, and education courses can be performed in units of modifiers, so that efficient courses can be provided. In addition, it is possible to design a system that saves labor.
In addition, by allowing the corrector (terminal) that first points out the character of the voice recognition error portion to perform the correction work, efficiency of the correction work can be realized.
[0077]
According to the second aspect of the present invention, by connecting the speech recognition error correction device and the terminal via a network, it is possible to flexibly cope with a work place, and it is possible to design a simplified system. Therefore, construction of an efficient speech recognition error correction system can be realized.
[0078]
According to the third and fourth aspects of the present invention, by changing the character attribute of a character in a speech recognition error portion (corrected character range) and presenting it to each terminal, the status of the pointing work or the correction work situation can be improved. Clarification can be achieved. Therefore, the correction work can be made more efficient.
[0079]
According to the fifth aspect of the present invention, at least one of the display width and the background color can be set for the display area of the text data as the speech recognition result based on a request from the terminal user. Thus, the display area of the text data becomes easy to see. Therefore, the visibility of the display screen can be improved.
[0080]
According to the invention described in claim 6, unnecessary characters are inserted into the text data as a result of the voice recognition, and the unnecessary characters are deleted by the corrector. Even in the strong case, the concentration of the corrector can be maintained. Therefore, the correction work can be maintained with high accuracy.
[Brief description of the drawings]
FIG. 1 is a schematic diagram illustrating a configuration of a speech recognition error correction system according to an embodiment of the present invention.
FIG. 2 is a block diagram illustrating a configuration of a speech recognition error correction device according to an embodiment of the present invention.
FIG. 3 is a diagram illustrating an example of an operation of an error text data protection unit of the text data correction unit according to one embodiment of the present invention.
FIG. 4 is a diagram illustrating an example of an operation of a character attribute changing unit of the text data correcting unit according to the embodiment of the present invention.
FIG. 5A is a diagram illustrating an example of an operation of an unnecessary character insertion unit of the text data correction unit according to an embodiment of the present invention.
(B) It is a figure for explaining an example of operation of the display field setting part of the text data correction part concerning one embodiment of the present invention.
FIG. 6 is a sequential chart for explaining an example of the operation of the speech recognition error correction device according to one embodiment of the present invention.
[Explanation of symbols]
1 Speech recognition error correction device
2 Voice recognition device
3 Audio playback unit
5 Voice receiver
7 Voice storage unit
9 Audio data transmission section
11 Voice presentation speed variable section
13 Text data correction unit
15 Character data receiving unit (data receiving means)
17 Correction information receiving unit (data receiving means)
19 Modified terminal operation determination unit
21 Character presentation speed variable section (presentation means)
23 Character attribute change unit (presentation means)
23a Designated character attribute change unit (pointed character attribute change function)
23b Modified character attribute change unit (Modified character attribute change function)
23c Character part of speech attribute changing section (character part of speech attribute changing function)
25 Character attribute information transmission unit (output means)
27 Display area setting section (display area setting function)
29 Unnecessary character insertion section (unnecessary character insertion function)
31 presentation information transmission unit (output means)
33 Text Data Protection Unit (Text Data Protection Means)
33a Modified guard terminal determination unit
33b Modified guard text judgment unit
35 String Integration Unit
37 screen display information transmission unit (output means)
39 Subtitle output component (output means)
41 (41A-41C) Correction terminal, pointing data transmission terminal (terminal)
43 Caption sending device
45 Television
AR text data display area
AF correction character input frame

Claims (8)

A speech recognition error which receives speech data output from a speech recognition device and which is a speech recognition result and text data as a speech recognition result, and corrects a speech recognition error included in the text data by a plurality of terminals. A correction device,
Presenting means for presenting the voice data and the text data to the terminal, and pointing data indicating a corrected character range indicated and represented by the terminal with respect to the text data presented to the terminal by the presenting means; Data receiving means for receiving corrected text data obtained by correcting text data in a corrected character range;
Based on the indication data received by the data receiving means, the corrected character range is indicated by a terminal other than the indication data transmission terminal until the correction by the indication data transmission terminal that is the terminal that transmitted the indication data is completed. Text data protection measures to protect against
An output unit that corrects the text data based on the corrected text data received by the data receiving unit and outputs the corrected text data to each of the terminals. The apparatus according to claim 1, wherein the plurality of terminals are connected via a network. The presenting means, for the characters in the correction character range, by changing the attribute of the character including at least one of the display color, display size, character type, to a terminal other than the indication data transmission terminal Pointed character attribute change function that also indicates the change of the character attribute,
When the corrected character range is being corrected by the indication data transmitting terminal, a corrected character attribute that has a character attribute different from the character attribute changed by the specified character attribute change function in the corrected character range being corrected. The speech recognition error correction device according to claim 1, further comprising a change function. The presenting means is a character part-of-speech attribute change that changes a character attribute including at least one of a display color, a display size, and a character type for at least particles of the part of speech related to the text data as the speech recognition result. The speech recognition error correction device according to claim 1 or 2, further comprising a function. The said presentation means is provided with the display area setting function which can arbitrarily set at least one of the display width and background color of the display area of the text data which is the said speech recognition result presented to the terminal. The speech recognition error correction device according to claim 1 or 2. 3. The speech recognition according to claim 1, wherein the presentation unit has an unnecessary character insertion function of inserting an unnecessary character into text data that is the speech recognition result presented to the terminal. Error correction device. A speech recognition error which receives speech data output from a speech recognition device and which is a speech recognition result and text data as a speech recognition result, and corrects a speech recognition error included in the text data by a plurality of terminals. How to fix it,
A presentation step of presenting the voice data and the text data to the terminal;
For the text data presented to the terminal by the presenting step, data reception for receiving indication data indicating a corrected character range indicated by the terminal and corrected text data obtained by correcting the text data of the corrected character range with respect to the text data presented to the terminal. Steps and
Based on the indication data received in this data receiving step, the corrected character range is indicated by a terminal other than the indication data transmission terminal until the correction by the indication data transmission terminal that is the terminal that transmitted the indication data is completed. Text data protection steps to protect against
An output step of correcting the text data based on the corrected text data received in the data receiving step and outputting the corrected text data to each of the terminals. A speech recognition error which receives speech data output from a speech recognition device and which is a speech recognition result and text data as a speech recognition result, and corrects a speech recognition error included in the text data by a plurality of terminals. Correction device,
Presentation means for presenting the voice data and the text data to the terminal;
Data reception for receiving, with respect to the text data presented to the terminal by the presenting means, indication data indicating a corrected character range indicated by the terminal and corrected text data obtained by correcting the text data of the corrected character range. means,
Based on the indication data received by the data receiving means, the corrected character range is indicated by a terminal other than the indication data transmission terminal until the correction by the indication data transmission terminal that is the terminal that transmitted the indication data is completed. Text data protection means to protect against
A speech recognition error correction program, which functions as output means for correcting the text data based on the corrected text data received by the data receiving means and outputting the corrected text data to each of the terminals.

Images