JP2019144310A - Information processor, information processing system, control method and program - Google Patents

Abstract

To provide technique for determining priority of proofreading in a character string being a result of voice recognition and allowing a proof-reader to perform proofreading.SOLUTION: An information processor comprising storage means for storing a text at every speech obtained by voice-recognizing voice data by a plurality of speeches and a first certainty factor being likelihood of a voice recognition result with respect to the text comprises: specification means for specifying the text at every speech to be proofread on the basis of the first certainty factor; display control means for allowing the specified text at every speech to be proofread to be identified and displayed; and update means for receiving proofreading with respect to the text for every speech and updating the text at every speech receiving proofreading.SELECTED DRAWING: Figure 3

Description

  The present invention relates to a technology for assisting in correcting an error in a speech recognition result, and presents a priority for proofreading from an important part when proofreading a character string having a recognition error. It relates to a technology that improves the accuracy of information security for readers who see the results.

  Conventionally, research and development of speech recognition for converting human speech into a character string has been performed. As actual applications, it has been used for purposes such as adding subtitles to TV broadcasts and deaf people understanding the speech of others.

  Particularly in recent years, practical recognition accuracy has been realized due to the progress of machine learning. However, sufficient recognition accuracy is not yet achieved, and it is necessary to pay particular attention to how the speaker speaks. For example, the recognition accuracy varies greatly depending on the distance between the microphone and the mouth, the utterance clarity, and the like.

  For this reason, software for correctly correcting the character string of the speech recognition result on a personal computer or the like has also been realized.

  Originally, there was a service to provide deaf people by listening to the utterer's utterances and inputting them into a personal computer as a “computer summary writing”. However, it is possible to support the input work by the advent of software using speech recognition technology. It is becoming possible.

  Such software is generally divided at the timing when the speech is interrupted, and the speech recognition result (character string) is also displayed on the screen in time series in the unit of the division. Those character strings will be corrected by a proofreader who has been trained as a PC summary writing.

  However, since the utterance is usually faster than the operation of correcting the recognition result, that is, the keyboard input operation in the information processing apparatus, the burden on the proofreader who performs the correction operation has not yet been sufficiently reduced.

  Patent Document 1 provides a speech recognition result editing apparatus that supports the work of a proofreader who corrects erroneous recognition of speech recognition correctly.

  In the technique of Patent Literature 1, an utterance input from a microphone is converted into a character string to which reliability is given for each word by a voice recognition unit. At this time, not only the word with the highest reliability but also a word satisfying a predetermined condition, for example, a word having a certainty of a certain value or more is converted into a character string and included as a speech recognition result set. The correct recognition result can be selected and corrected from the correct answer candidates (paragraph 0013 of FIG. 1, FIG. 8).

JP 2017-040856 A

  However, since a single utterance includes a plurality of words, with the technique of Patent Document 1, when the number of words is large, the number of recognition results is enormous, and it is difficult to display on the display device in an easy-to-understand manner. In particular, in the case of the above-mentioned personal computer summary writing, since the correction is made in an easy-to-understand manner by judging from the context, the character string of the utterance before the utterance being corrected may be confirmed, but there may be no area. That is, simply displaying the recognition result candidates may not provide efficient support for the proofreader.

  If there are not enough proofreaders, all speech recognition errors may not be corrected. In this case, it is necessary to preferentially determine the part to be corrected. However, in the technique described in Patent Document 1, the speech recognition result that has not been corrected and a fixed time has passed is corrected regardless of its importance. It is only determined that it is not performed (timeout process described in paragraph 0022 of FIG. 5, Patent Document 1), and even if important information is passed, it is discarded or displayed without modification after a certain period of time. is there.

  Moreover, although it is ideal that the character strings corresponding to all utterances are corrected by the proofreader, there are cases where it is impossible in practice. When it is not possible, it is necessary to provide information that is easy to understand as much as possible to a person who sees the correction result (for example, a deaf person) by giving the proofreader correction with an appropriate priority.

  In view of the above problems, an object of the present invention is to provide a technique for judging a correction priority order and correcting a proofreader among character strings as a result of speech recognition.

  An information processing apparatus comprising storage means for storing a text for each utterance obtained by voice recognition of voice data of a plurality of utterances and a first certainty factor that is a likelihood of a voice recognition result for the text, Specification means for specifying text for each utterance to be corrected based on the first certainty factor, display control means for identifying and displaying the text for each specified utterance to be corrected, and correction for the text for each utterance And updating means for updating the text for each utterance for which the correction has been accepted.

  According to the present invention, in view of the above-described problem, it is possible to provide a technique for judging a correction priority order and correcting a proofreader among character strings as a result of speech recognition.

It is a figure which shows an example of the system configuration | structure which concerns on embodiment of this invention. It is a block diagram which shows an example of the hardware constitutions of the information processing apparatus which concerns on embodiment of this invention. It is a figure which shows an example of the function structure which concerns on embodiment of this invention. It is a figure which shows an example of the screen which displays the speech recognition result which concerns on embodiment of this invention. It is a figure which shows an example of the process from the voice person input which concerns on embodiment of this invention to distribution of calibration. It is a figure which shows an example of the data format of the recognition result which concerns on embodiment of this invention, and the certainty factor of a recognition result. It is a figure which shows an example of the flowchart explaining the process to the prioritization for the analysis of the speech recognition result which concerns on embodiment of this invention, and calibration. It is a figure which shows an example of the flowchart explaining the process of prioritization for the calibration which concerns on embodiment of this invention. It is a figure for showing an example of the memory | storage part which memorize | stores the information used for the process of the priority which concerns on embodiment of this invention. It is a figure for showing an example of the result of having recalculated the certainty degree of the voice recognition result concerning the embodiment of the present invention. It is a figure for showing an example of the user interface which displays the voice recognition result concerning the embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 1 is a diagram illustrating an example of a system configuration according to an embodiment of the present invention.
<System configuration example 1>

  The system according to the embodiment of the present invention includes a speech recognition server 101 and an information processing terminal 102 (referred to as a speaker 102a, a proofreader 102b, and a reader 102c). The user inputs voice through the microphone 104 connected to the information processing terminal 102a. The information processing terminal 102a transmits the voice to the voice recognition server 101, converts it into a character string, sends it to the information processing terminals 102a to 102c, displays the information on the information processing terminals 102a to 102c, and presents it to the user. That is, the information processing terminals 102a to 102c may have both input and output of voice input and character string output. The information processing terminal 102 output here may serve both as a reader 102c and a proofreader 102b described later, or may be a dedicated information processing terminal. Further, output is performed on a display device connected to the information processing terminal 102, but a configuration using a projector or the like is also a system configuration according to the embodiment of the present invention. If a projector is used, only one information processing terminal 102 is provided for a speaker, and all readers may read a character string of a speech recognition result displayed on a screen from a projector connected to the information processing terminal 102a. In that case, the misrecognition may be corrected as a proofreader by the utterer himself or another user directly at the information processing terminal 102a for the speaker.

Furthermore, the voice recognition server 101 may exist on the cloud, and in this case, the user of the present system may be in the form of a cloud service for functions on the voice recognition server 101 described later. Even in the form of using these services, the system configuration according to the embodiment of the present invention is adopted.
<System configuration example 2>

The information processing terminals 102a to 102c described in the configuration example 1 have both input and output, but may be separated from input only and output only.
<System configuration example 3>

  The voice recognition server 101 and the information processing terminals 102a to 102c may be the same housing. That is, software capable of voice recognition may be installed in one of the information processing terminals 102 a to 102 c in FIG. 1 and may also serve as the voice recognition server 101.

  FIG. 2 is a block diagram showing an example of a hardware configuration applicable to the voice recognition server 101 and the information processing terminals 102a to 102c according to the embodiment of the present invention.

  As shown in FIG. 2, the speech recognition server 101 and the information processing terminals 102 a to 102 c include a CPU (Central Processing Unit) 201, a RAM (Random Access Memory) 202, a ROM (Read Only Memory) 203, and the like via a system bus 204. A configuration in which an input controller 205, a video controller 206, a memory controller 207, a communication I / F controller 208, and the like are connected is adopted. The CPU 201 comprehensively controls each device and controller connected to the system bus 204.

  Further, the ROM 203 or the external memory 211 will be described later, which is necessary for realizing the functions executed by each server or each PC, such as BIOS (Basic Input / Output System) and OS (Operating System) which are control programs of the CPU 201. Various programs are stored. Further, information necessary for carrying out the present invention is stored. The external memory may be a database.

  The RAM 202 functions as a main memory, work area, and the like for the CPU 201. The CPU 201 implements various operations by loading a program or the like necessary for executing the processing from the ROM 203 or the external memory 211 to the RAM 202 and executing the loaded program.

  The input controller 205 controls input from a keyboard (KB) 209 or a pointing device such as a mouse (not shown).

  The video controller 206 controls display on a display device such as the display 210. The display device may be a display device such as a liquid crystal display. These are used by the administrator as needed.

  The memory controller 207 is an external storage device (hard disk (HD)), flexible disk (FD), or PCMCIA (Personal Computer) that stores a boot program, various applications, font data, user files, editing files, various data, and the like. Controls access to an external memory 211 such as a Compact Flash (registered trademark) memory connected to a Memory Card International Association (Card Memory) card slot via an adapter.

  The communication I / F controller 208 connects and communicates with an external device via a network, and executes communication control processing on the network. For example, communication using TCP / IP (Transmission Control Protocol / Internet Protocol) is possible.

  Note that the CPU 201 can display on the display 210 by executing an outline font rasterization process on a display information area in the RAM 202, for example. Further, the CPU 201 enables a user instruction using a mouse cursor (not shown) on the display 210.

Various programs to be described later for realizing the present invention are recorded in the external memory 211 and executed by the CPU 201 by being loaded into the RAM 202 as necessary.
FIG. 3 is a diagram illustrating an example of a functional configuration according to the embodiment of the present invention.

  The information processing terminal 102 may have the functions of the speaker 102a, the proofreader 102b, and the reader 103b in separate terminals or in a common terminal. Explained.

  The voice acquisition unit 311 inputs speech uttered by the speaker's voice as voice data from a microphone incorporated in the information processing terminal 102 or connected to the voice recognition server 101 by the voice data transmission unit 312.

  The voice recognition server 101 passes the voice data received by the voice data receiving unit 321 to the voice recognition unit 322, converts the voice data into a character string, and the recognition result transmission unit 323 sends the recognition result to the information processing terminal 102. Send back as. Further, the recognition result management unit 324 stores the above-described recognition result in the recognition result storage unit 320.

  The information processing terminal 102 receives the character string at the recognition result receiving unit 313 and displays it on the display unit 314 to present it to the reader (user of the information processing terminal 102).

  The priority determination unit 325 is stored in the recognition result storage unit 320 so that a proofreader for proofreading a speech recognition error using the information processing terminal 102 can identify a character string to be proofread preferentially. Give priority to recognition results.

  The prioritized character strings are transmitted to the information processing terminal 102, and the display unit 314 of the information processing terminal 102 displays the priority order to be proofread by the proofreader as described above. The recognition result proofreading unit 315 provides a function for proofreading the recognition result error by the proofreader editing the character string.

  The calibration result is transmitted to the speech recognition server 101 by the calibration result transmission unit 316 of the information processing terminal 102, received by the calibration result reception unit 326 of the speech recognition server 101, and stored in the recognition result storage unit 320. Update the result.

The updated recognition result is distributed to the information processing terminal 102 other than the information processing terminal 102 used for proofreading by the proofreader by the proofreading result distributing unit 327 so that the reader can see the proofreading result. Presented.
FIG. 4 is a diagram showing an example of a screen that displays a speech recognition result according to the embodiment of the present invention.

  The utterance example 400 is an utterance example of a speaker in a conference or lecture. There is no need to specify a single speaker. For example, in a meeting, there is a possibility of speaking in addition to the chairperson. In lectures, there are utterances by the presenter and questioners in addition to the speakers. Also good.

  In the utterance example 400, the utterance example 400 is divided into A to K. These are breaks of the utterance of the speaker. For example, it shows a case where there is a blank (silenced state) for a certain time in the utterance.

  On the other hand, the voice recognition result display screen 401 is also divided corresponding to the above A to K (a plurality of display frames 404A to K), but these are silenced by the voice recognition unit 322 of the voice recognition server 101. The character string of the recognition result is delimited by recognizing the state of. These are separated and stored in the recognition result storage unit 320, and the display unit 314 of the information processing terminal 102 displays them in a manner that is easy to understand for the reader. This is an example, and 404 does not necessarily have to be divided, but is merely a design matter. It suffices if the priority order for proofreading described later is displayed so as to be recognizable.

  A start button 402 is pressed when the speech recognition server 101 recognizes an utterance. The system configuration diagram (FIG. 1) shows a diagram of a plurality of information processing terminals 102 and microphones connected to them, and designates which information processing terminal 102 is speaking to a microphone connected to the information processing terminals 102 Is for. An utterance may be input to only one information processing terminal 102, or an utterance may be input to a plurality of information processing terminals 102 at the same time, depending on the system design. Further, there may be an end button 403 for notifying the information processing terminal 102 that no utterance has been input in response to the start button 402.

  Among 404A to 404K, 404A to J indicate a state in which the "blank for a certain period of time (silent state)" has passed. On the other hand, 404K is a state in which the speech recognition unit 322 still determines that the speaker's utterance is continuing as the recognition result is being output. In the figure, a part of the utterance is already recognized, and the part is displayed. However, after the break appears, the speech recognition results of the utterance may be displayed together.

  FIG. 5 is a diagram illustrating an example of processing from voice person input to proofreading distribution according to the embodiment of the present invention. Each step of the flowchart of FIG. 5 is executed by the CPU 201 on the speech recognition server 101 and the CPU 201 on the information processing terminals 102a to 102c.

  In step S501, a speaker's utterance is received through a microphone or the like connected to the information processing terminal 102a, and converted into voice data.

  In step S502, the information processing terminal 102a transmits the voice data to the voice recognition server 101, and the voice recognition server 101 receives the voice data in step S503.

  In step S504, the voice recognition server 101 converts the utterance of the speaker in the voice data into a character string by voice recognition. The character string of the recognition result is divided in units of utterances as described above, but is further divided in such a way that it can be identified in linguistic units such as morphemes. As a result of the speech recognition, not only a character string but also a certainty is given to the probability that the speech recognition unit 322 estimates the recognition result to be correct. In addition, in the case where the morpheme is divided into linguistic units, a certainty factor and a detailed part of speech may be assigned to each morpheme as a tag. Participation by morphological analysis will be explained with reference to an example in FIG. 10. The school grammar learned at school is roughly “proprietary nouns”, but in information processing, for example, proper nouns are “person names” and “place names”. And so on. Morphological analysis and speech recognition are well-known techniques and will not be described in detail.

  In step S505, the speech recognition server 101 transmits the character string that is the conversion result in step S504 to the information processing terminal 102. When a plurality of information processing terminals 102 are connected in the system, the character string is transmitted not only to the information processing terminal 102a that inputs the utterance but also to all the information processing terminals 102. The speaker himself / herself may also transmit to the information processing terminal 102a used by the speaker and receiving the voice data in order to confirm the speech recognition result. The information processing terminal 102 receives the character string in step S506.

  In step S507, the speech recognition server 101 stores the speech recognition result in the recognition result storage unit 320. The format in which the recognition result is stored will be described in detail with reference to FIG.

  FIG. 6 is a diagram showing an example of a data format of the recognition result and the certainty factor of the recognition result according to the embodiment of the present invention. As an example, it is assumed that the recognition result is stored in the structure of the recognition result information 600.

  601A to J are data corresponding to the utterances A to J in FIG. The character string that is the result of the speech recognition unit 322 is stored in the recognized character string 603 corresponding to the utterance delimiter as described above. 602A to J are certainty degrees corresponding to the utterances 601A to 601J. Each recognition result includes a recognition character string 603, a morpheme notation 604, which will be described later, and a certainty factor 605 of the recognition result of each morpheme.

  Also, a dark background in the frame of the morpheme notation 604 (606, etc.) will be described later with reference to FIGS. 8 to 10. However, if the speech recognition result is misrecognized, it is difficult for the reader to understand. Used to determine which part to calibrate.

  In step S508, the speech recognition server 101 determines the priority of calibration for managing one or more data stored in the recognition result storage unit 320 in the process up to the above-described step S507 when a new utterance is input. Manage. That is, the recognition result information 600 in FIG. 6 is managed. These processes will be described in detail with reference to FIGS.

  Asynchronously with the processing in the speech recognition server 101, the proofreader information processing terminal 102b presents the character string received in step S506 to the proofreader on the display device of the information processing terminal 102b. In step S509, Accept proofreaders' proofreading work. The proofreading work of the proofreader is to edit the character string corresponding to the utterance while following the identifiable priority order displayed on the display device of the information processing terminal 102b. A screen in a state where calibration is performed will be described later with reference to FIG. Further, when the calibration work is started in step S509, the voice recognition server 101 is notified of this, and the correction state of the data stored in the recognition result storage unit 320 is changed to “under calibration”.

  In step S510, a character string as a result of completion of the above-described calibration is transmitted from the information processing terminal 102b. In step S511, the speech recognition server 101 receives the result and stores it in the recognition result storage unit 320. Update the data. At that time, the correction state is changed to “completed” and the necessity of correction is changed to “unnecessary”.

  In step S512, the proofreading result distribution unit 327 of the speech recognition server 101 transmits to the information processing terminal 102 the character string that has been proofread, that is, the character string in which the misrecognition portion in speech recognition has been calibrated.

  The correct information processing terminal 102b for the proofreader who corrects the error already displays the correct character string at the time of the correction, but transmits the correct character string to the information processing terminal 102b, that is, to itself as a design matter. May be. In the flowchart of FIG. 5, the character string that has been proofread is once distributed to the information processing terminal 102 via the voice recognition server 101, but other information processing is directly performed from the information processing terminal 102 b for proofreading. You may distribute to the terminal 102. This difference is merely a design matter, and the case of direct distribution is included in the scope of the claims of the present invention.

  In step S513, the information processing terminal 102 receives the calibrated character string, and changes the “character string including erroneous recognition” already displayed on the display device of the information processing terminal 102 to “calibrated character string”. replace.

  Note that the display frames 404A to 404K in FIG. 4 may be different editing targets for each utterance, or may be a single editing target. In addition, the display frame 404 being configured by one information processing terminal 102b may not be calibrated by another information processing terminal 102b so that a plurality of proofreaders do not calibrate one display frame 404 at the same time. . Further, the display frame 404 at the bottom of FIG. 4 may be configured so that it cannot be proofread because the display of the character string that is not divided into voice recognition continues. These are only design matters.

  FIG. 7 is a diagram illustrating an example of a flowchart for explaining processing (step S508 in FIG. 5) up to prioritization for analysis and calibration of a speech recognition result according to the embodiment of the present invention. Each step of the flowchart of FIG. 7 is executed by the CPU 201 on the voice recognition server 101.

  In step S701, it is checked whether or not the recognition result of the speech data of the new utterance has been registered in the recognition result storage unit 520. More specifically, assuming that up to 601J in FIG. 6 existed in the previous check, it is checked whether or not the next 601K has been newly added. If registered (in the case of “Yes”), the process proceeds to step S702. If not registered (in the case of “No”), the process proceeds to step S704.

  In step S702, morphological analysis is performed on the newly added character string of the speech recognition result. As shown in the example of FIG. 10, the part of speech is given by dividing the character string by the process of step S702. As a result, a morpheme string is generated. However, if the speech recognition result itself has a part of speech by morphological analysis, step S702 is unnecessary and is omitted.

  In step S703, an individual name is extracted from the morpheme string. The technique for extracting an individual name is a well-known technique disclosed in Japanese Patent Application Laid-Open No. 2002-288190 and the like, and thus detailed description thereof is omitted.

  In step S704, a priority order to be calibrated is set for a character string (for example, 601A to J in FIG. 6) that is a recognition result that has not been calibrated. Details will be described later with reference to FIGS.

  In step S705, when the execution of the voice recognition system continues (in the case of “Yes”), the process returns to step S701. When the execution of the speech recognition system is finished (in the case of “No”), the process of the flowchart of FIG. 7 is completed, and the process returns to the process of the flowchart of FIG. That is, the process returns to the state after step S508 in FIG.

  FIG. 8 is a diagram showing an example of a flowchart for explaining the prioritization processing (step S704 in FIG. 7) according to the embodiment of the present invention. Each step of the flowchart of FIG. 8 is executed by the CPU 201 on the voice recognition server 101.

  Steps S801 to S808 are repetitive processing on the result stored in the recognition result storage unit, that is, the result of speech recognition based on all speech data (for example, 10 speech data in the case of 601A to J in FIG. 6).

  In step S802, attention is focused on one speech recognition result. Specifically, attention is paid to one of the 601A to J in order from the top.

  In step S803, it is determined whether or not it is necessary to determine the priority order of the speech recognition result under consideration. The process branches depending on whether calibration has already been performed or whether the condition described in the post-speech progress condition 901 in FIG. 9 is satisfied. Since this determination is an OR condition for two types of determinations, if either condition is satisfied, “Yes” is determined, and the process proceeds to step S804. If neither condition is satisfied, “No” is determined, and the process proceeds to step S805.

  Whether or not the two conditions are already calibrated will be described in detail with reference to FIG. 10 (recognition state is given as an example of a part of the recognition results in FIG. 6). When the flowchart in FIG. 8 (that is, step S704 in FIG. 7) is executed for the first time after voice recognition of a certain segment of speech, the “necessity of correction” in FIG. 10 of the speech has not yet been determined. The condition is not satisfied because it is a blank state with no description (“No”). The recognition result that has already been calibrated satisfies the condition (“Yes”) since “correction necessity” in FIG. 10 is set to “unnecessary” in FIG. However, this part is a design matter, and the recognition result once calibrated may be reassigned. In that case, it is not determined as “unnecessary” in S510.

  A case where the post-speech progress condition 901 out of the two is used as a condition will be described. The intent of this condition is to determine that a certain period of time has passed after the utterance is completed and that it is not useful to calibrate retrospectively. . Specifically, description will be made using three examples described in 901 of FIG.

  The post-speech progress condition 901 describes how to determine that a certain time has passed since the speech was made. The conditions shown in FIG. 9 are merely examples, and any method for determining the passage of time, other than these three methods, is included in the present invention. One example will be explained at a time.

  Example 1 is to measure the actual time from the segment where the utterance is considered to be completed in each utterance such as A to J in the utterance example 400 of FIG. In the example, when 180 seconds or more have passed since completion, the condition is that calibration is unnecessary. The elapsed time is stored in the “Elapsed Time” field in the example of FIG.

  Example 2 counts by the number of characters, not the time. If the utterance is considered to have been completed and separated, if more than 500 characters of the following utterance are presented as recognition results, calibration is not necessary at that time. And Referring to 603 in FIG. 6, if the total number of characters after 603B reaches 500 characters after 603A, the priority of the speech of 601A is not calculated and calibration is unnecessary.

  Example 3 is determined based on how the reader sees it. The character string of the speech recognition result is usually not displayed on the information processing terminal 102 of the reader as time passes. For example, the speech recognition result display screen 401 in FIG. 4 and FIG. 11 is displayed in chronological order from the top in utterance breaks, and when the screen is full, the latest one is added to the bottom row, so that the top row (the oldest utterance) Can be considered as a user interface that scrolls and disappears upward (for example, a portion within a dotted line 1101 in FIG. 11).

  Even if there is a different method that is not in chronological order, that is, something that remains on the screen / disappears on the screen regardless of the old and new, even if you correct the error of what has disappeared, any reader Because it cannot be read, it is meaningless. Therefore, it can be considered that calibration is unnecessary.

  Three examples are given here, but other methods may be used. Moreover, these combination conditions (AND condition, OR condition) may be sufficient.

  In step S804, in order to eliminate the need for calibration, “correction necessity” of the information stored in FIG. 10 is set to “unnecessary”.

  In step S805, “necessary for correction” in the calibration status is set as “necessary” when it is necessary to calibrate, and “corrected state” is set as “not yet” as not yet calibrated.

  Incidentally, as already described, at the stage where calibration is started in step S509 of FIG. 5, the “correction state” is “under calibration”, the calibration is completed, and the calibration result is transmitted to the voice recognition server 101, step S511. The correction state is changed to “completed” and the necessity of correction is changed to “unnecessary”.

  Next, in step S806, for example, according to 902 in FIG. Although three conditions are described in 902 as an example, it is not limited to these conditions.

  For example, in Example 1 of 902, it is determined whether the speech recognition result under attention includes a morpheme or individual name of a part of speech to be confirmed. For example, “B” in FIG. 10 includes “numerical”, and this is registered as a part of speech to be confirmed in 903 in FIG. In general, a specific pattern including a number or a numerical value is a company's sales, contract amount, date, etc., so if there is an error, important information for the reader is not guaranteed. Further, E in FIG. 10 includes a numerical expression (1002) that is a result of the individual name extraction. It is an individual name that needs to be confirmed that there is no error even when it includes a specific person, organization, and numerical expression calibrated from a plurality of morphemes (904 in FIG. 9).

  As the second example of 902, when many morphemes with particularly low confidence are included in the speech recognition result, the third example is a case where the certainty of the recognition result of the entire utterance is low. Yes. When the certainty of recognition is low, there is a high possibility that many misrecognized morphemes are included, and therefore the priority of calibration is high in a sense different from Example 1 in which there is important information individually.

  The certainty factor is recalculated according to processing such as morphological analysis / individual name extraction and the rules described in 902 and the like. The certainty factor recalculation method is described in the certainty factor recalculation method 905 as an example. That is, if important information is included in the above-described processing, the priority of calibration is changed by changing the certainty of the recognition result. For example, it is described how to recalculate the certainty factor when there is a word registered in the part of speech to be confirmed 903 and information specified by the individual name extraction condition 904 (Example 1 of 905, Example 1) 2).

  Although the flowchart morphological analysis process of FIG. 8 is not described here, the speech recognition result itself is often divided into morpheme units, and the part of speech may be included as information of the speech recognition result. . If they are not included, part-of-speech assignment by morphological analysis or other methods (such as using a dictionary) may be performed separately.

  The same applies to individual name extraction. It may be included as a part of the embodiment of the present invention, or may include any result including an individual name extracted on the speech recognition side as a speech recognition result.

  In step S807, a calculation is performed to change the configuration priority according to the time after the utterance ends. In the determination of step S803 and the example 1 of 901, the calibration is not required if the fixed time has elapsed, but here it corresponds to the recognition result before the fixed time has elapsed. That is, for example, a certain period of time has not elapsed (Example 1 of 901), and the recognition result still displayed on the screen (Example 3 of 901) is approaching a state where calibration is “unnecessary”. As the time limit left for calibration is small, it is necessary to raise the priority and perform calibration. The expression of Example 3 of 905 decreases the certainty of the recognition result as time passes.

  In step S809, in response to the result of the recalculation of the certainty factor as described above, the information that the calibration is “necessary” is sorted by the certainty factor, and the lower the certainty factor, the information is preferentially calibrated. This is presented on the display device of the processing terminal 102.

  This completes the description of the processing according to the flowchart of FIG. Here, the certainty factor is changed according to a certain rule, but the certainty factor is not necessarily changed. For example, another numerical value (such as a minus score) for storing how much “deduction” has been used may be used. Changing the certainty factor is merely an example and a design matter.

  The priority order to be calibrated is determined by performing the processing of FIG. 8, and FIG. 10 shows an example in which the recognition result E is priority order 1 and the recognition result I is priority order 2, for example. . The proofreader determines the priority order for proofreading based on this identifiable information. Or you may control so that it can edit only from a thing with high priority.

  FIG. 11 is a diagram illustrating an example of a user interface that displays a speech recognition result according to the embodiment of the present invention. This is essentially the same as FIG. 4 except for the following points.

  1103e displays “medium” (calibrating) indicating that one of the proofreaders is proofreading the recognition result. Priorities 1 to 4 are assigned to 1103g to 1103j. Thus, the proofreader can identify the priority order to be proofread.

Further, 1103k is displayed as “current” because the speech recognition result currently being spoken is recognized halfway and the result is displayed. Even if the display frame can be proofread, it may be controlled so that it cannot be proofed until the utterance is divided and the next 1103 l is displayed.
This completes the description of the present invention using the drawings.

  It should be noted that the configuration and contents of the various data described above are not limited to this, and it goes without saying that the various data and configurations are configured according to the application and purpose.

  Although several embodiments have been described above, the present invention can take an embodiment as, for example, a system, apparatus, method, computer program, or recording medium, and more specifically, a plurality of devices. The present invention may be applied to a system configured from the above, or may be applied to an apparatus including a single device.

  Further, the computer program in the present invention is a computer program that can execute the processing method of the flowcharts shown in FIGS. 5, 7, and 8, and the storage medium of the present invention is the process in FIG. 5, FIG. 7, and FIG. A computer program capable of executing the method is stored. Note that the computer program in the present invention may be a computer program for each processing method of each apparatus in FIGS. 5, 7, and 8.

  As described above, a recording medium in which a computer program for realizing the functions of the above-described embodiments is recorded is supplied to the system or apparatus, and the computer (or CPU or MPU) of the system or apparatus is stored in the recording medium. It goes without saying that the object of the present invention can also be achieved by reading and executing a program.

  In this case, the computer program itself read from the recording medium realizes the novel function of the present invention, and the recording medium storing the computer program constitutes the present invention.

  As a recording medium for supplying a computer program, for example, a flexible disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a DVD-ROM, a magnetic tape, a nonvolatile memory card, a ROM, an EEPROM, Silicon disks, solid state drives, etc. can be used.

  Further, by executing the computer program read by the computer, not only the functions of the above-described embodiments are realized, but also an OS (operating system) running on the computer based on the instructions of the computer program. It goes without saying that a case where the function of the above-described embodiment is realized by performing part or all of the actual processing and the processing is included.

  Furthermore, after the computer program read from the recording medium is written in a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer, the function is based on the instructions of the computer program code. It goes without saying that the CPU or the like provided in the expansion board or the function expansion unit performs part or all of the actual processing and the functions of the above-described embodiments are realized by the processing.

  Further, the present invention may be applied to a system composed of a plurality of devices or an apparatus composed of a single device. It goes without saying that the present invention can also be applied to a case where the present invention is achieved by supplying a computer program to a system or apparatus. In this case, by reading the recording medium storing the computer program for achieving the present invention into the system or apparatus, the system or apparatus can enjoy the effects of the present invention.

Furthermore, by downloading and reading out a computer program for achieving the present invention from a server, database, etc. on a network using a communication program, the system or apparatus can enjoy the effects of the present invention.
In addition, all the structures which combined each embodiment mentioned above and its modification are also included in this invention.

101 voice recognition server 102 information processing terminal 320 recognition result storage unit 321 voice data reception unit 322 voice recognition unit 323 recognition result transmission unit 324 recognition result management unit 325 priority order determination unit 326 calibration result reception unit 327 calibration result distribution unit

Claims (9)

An information processing apparatus comprising storage means for storing a text for each utterance obtained by voice recognition of voice data of a plurality of utterances and a first certainty factor that is a likelihood of a voice recognition result for the text,
A specifying means for specifying text for each utterance to be corrected based on the first certainty factor;
Display control means for identifying and displaying the text for each utterance to be corrected,
An information processing apparatus comprising: an update unit configured to receive correction for the text for each utterance and update the text for each utterance for which the correction has been received. The storage means stores a morpheme obtained by analyzing the text for each utterance and a second certainty factor that is a probability of a speech recognition result for the morpheme,
The information processing apparatus according to claim 1, wherein the specifying unit specifies a text for each utterance to be corrected based on a second certainty factor of a morpheme included in the text for each utterance.   The information processing apparatus according to claim 2, wherein the specifying unit specifies a text for each utterance to be corrected based on whether or not a morpheme included in the text for each utterance has a predetermined part of speech. A determination means for determining whether the obtained morpheme is an individual name;
The information processing apparatus according to claim 2, wherein the specifying unit specifies the text for each utterance to be corrected depending on whether or not a morpheme included in the text for each utterance is an individual name. .   The display control means highlights the morpheme when the morpheme included in the text for each utterance has a predetermined part of speech or an individual name. The information processing apparatus described in 1.   The information processing apparatus according to claim 1, wherein the specifying unit specifies a text for each utterance to be corrected according to an elapsed time from the utterance. An information processing apparatus and a display device including storage means for storing a text for each utterance obtained by voice recognition of voice data of a plurality of utterances and a first certainty factor that is a likelihood of a voice recognition result for the text. An information processing system including:
A specifying means for specifying text for each utterance to be corrected based on the first certainty factor;
Display control means for causing the display device to identify and display the text for each specified utterance to be corrected;
An information processing system comprising: correction accepting means for accepting a correction to the text for each utterance in the display device. A control method for an information processing apparatus comprising storage means for storing text for each utterance obtained by speech recognition of speech data of a plurality of utterances and a first certainty factor that is the likelihood of a speech recognition result for the text. And
A specifying step of specifying a text for each utterance to be corrected based on the first certainty factor;
A display control step for causing the display control means to identify and display the text for each utterance to be corrected,
An update means comprises: an update step of accepting a modification to the text for each utterance and updating the text for each utterance that has accepted the modification. A program that can be executed in an information processing apparatus including a storage unit that stores text for each utterance obtained by voice recognition of voice data of a plurality of utterances and a first certainty factor that is a likelihood of a voice recognition result for the text Because
The information processing apparatus;
A specifying means for specifying text for each utterance to be corrected based on the first certainty factor;
Display control means for identifying and displaying the text for each utterance to be corrected,
A program for accepting a modification to the text for each utterance and to function as an updating unit for updating the text for each utterance for which the modification has been accepted.

Images