JP4020083B2 - Transcription text creation support system and program - Google Patents

Description

  The present invention relates to a transcribed text creation support system that supports a task of converting speech data into a sentence, and more particularly, to appropriately correct a recognition result obtained by performing speech recognition processing on speech information and create a final transcribed text. The present invention relates to a transcription text creation support system.

  Converting audio information included in content into text is called transcription or simply transcription, and has been practiced in various fields. For example, to search for content of video / audio content using keywords and to realize video search / distribution service that distributes video / audio from scenes that hit the search, the audio in the video / audio content is transcribed into text. And making it a search keyword. In addition, audio information is written as text as subtitle information for video content.

  The procedure for creating the transcript text is that the operator listens to the audio information of the content to be transcribed, transcribes the audio information using an input means such as a keyboard, and then visually reviews the transcribed text to make a transcription error. It is common to find and correct. However, since this kind of work requires a lot of effort, a system has been devised to improve the efficiency of the transcription work by automatically generating text from speech information to some extent using speech recognition technology. (For example, refer to Patent Document 1).

  However, in the system for improving the efficiency of transcription work using speech recognition technology, the operator visually detects and corrects errors in the speech recognition result text output from the system while listening to the actual speech information by ear. There is a problem that a work load is eventually applied, such as confirming which part is being worked when repeatedly confirming, and moving back and forth the error reproduction point.

  As one solution, Patent Document 2 automatically finds an error part by comparing actual speech information with speech information uttered by speech synthesis technology based on the recognition result text. The method of presenting is presented. As a result, the load on the operator is expected to be reduced to some extent. In practice, however, it is necessary to manually check the entire target content unless the accuracy of the point automatically pointed out to be incorrect is 100%. Therefore, there remains a problem that it does not fundamentally reduce the workload.

On the other hand, Non-Patent Document 1 discloses a recognition result list in which a speaker, time information, a speech recognition result, and a transcription text are displayed in one line for each sentence that is an utterance unit, and a high level in the recognition result list. The voice automatic playback window that displays the waveform of the voice information corresponding to the utterance unit that is displayed in light and the editing window of the transcription text are displayed on the screen, and the voice recognition result is interactively played while the voice is played back in the utterance unit. Transcript text support system that can be modified is published.
JP 2001-166790 A JP 2001-134276 A “Introduction of“ Transcription System ”by Voice Recognition”, [online], Advanced Media, Inc. [Search February 4, 2004], Internet <URL: http://www.advanced-media.co .jp / event / news / AmiVoice_Rewriter.pdf>

  According to the transcription text support system described in Non-Patent Document 1 described above, it is possible to interactively correct the speech recognition result for each utterance while reproducing the speech for each sentence that is the utterance. It is possible to create a transcribed text with a certain degree of efficiency. However, since the sound is played in sentence units, even if some words in the sentence are suspicious, it is necessary to listen to the sound from the beginning of the sentence and listen to the sound of the corresponding word part efficiently. And it is difficult to hear the voice of the word accurately. Such a problem becomes more prominent as the utterance sentence becomes longer. In addition, since the correspondence between the speech information and the corresponding speech recognition result and the transcription text is one sentence unit, which part of the speech information corresponds to each part of the speech recognition result and the transcription text. It is unknown and it is not easy to confirm whether all of the audio information has been written down without omission.

  The present invention was devised based on the recognition of the above technical problem, and its purpose is to provide a transcription text creation support capable of efficiently and accurately creating a transcription text of audio information included in content. To provide a system.

  The first transcript text creation support system of the present invention is obtained by performing speech recognition processing on speech information included in a storage device, a display device, an input device, an audio output device, and material content. For each division unit when the speech recognition result is divided into units smaller than sentences, the speech recognition result of the division unit, the original speech information, time information for identifying the speech information, and transcription of the initial state A synchronization information generating unit that associates text and stores it in the storage device, a speech recognition result for each division unit stored in the storage device, an image of the speech information that is the basis thereof, and the speech information The time information and the transcribed text are aligned and displayed in parallel on the screen of the display device with the time axis aligned, and a voice reproduction finger specifying a division unit on the screen by the input device. The voice information of the designated division unit is reproduced and output from the voice output device, and the storage is performed in response to the input operation of the transcription text designating the division unit on the screen by the input device. And a confirmation correction unit for updating the transcription text of the designated division unit in the apparatus.

  The second transcription text creation support system of the present invention is characterized in that the division unit is a word in the first transcription text creation support system.

  The third transcription text creation support system according to the present invention is characterized in that in the first transcription text creation support system, the image of the voice information is an image of a voice waveform generated from the voice information.

  According to a fourth transcription text creation support system of the present invention, in the first transcription text creation support system, the input operation of the transcription text specifying the division unit on the screen by the input device writes the recognition result. It is a copy operation for copying to a wake-up text.

  According to a fifth transcript text creation support system of the present invention, in the first transcript text creation support system, the confirmation correction unit displays the screen on the display device in response to a scroll instruction from the input device. The speech recognition result for each divided unit, the image of the original speech information, the time information for specifying the speech information, and the transcribed text are scrolled in synchronization with each other. To do.

  A first program of the present invention is a speech recognition result obtained by performing speech recognition processing on speech information included in material content on a computer including a storage device, a display device, an input device, and an audio output device. Is divided into units smaller than sentences, and for each division unit, the speech recognition result of the division unit, the original voice information, time information for identifying the voice information, and the initial transcript Synchronization information generating means for storing in the storage device in association with each other, a speech recognition result for each division unit stored in the storage device, an image of the original speech information, and time information for specifying the speech information; Transcript text is aligned and displayed in parallel on the screen of the display device with the time axis aligned, and responds to an audio playback instruction specifying the division unit on the screen by the input device The audio information of the designated division unit is reproduced and output from the audio output device, and the input unit of the transcription text designating the division unit on the screen by the input device is used in the storage device. It is made to function as a confirmation and correction means for updating the transcription text of the designated division unit.

  According to a second program of the present invention, in the first program, the division unit is a word.

  According to the present invention, it is possible to efficiently and accurately create a transcription text of audio information included in content.

  The first reason is that audio information can be played back in smaller units than sentences, so if there is any doubt about a part of a piece of voice information, only that part can be played back, and it can be heard efficiently and accurately. Secondly, since the speech information and the speech recognition result and the transcription text corresponding to the speech information are displayed in association with each unit smaller than the sentence, the speech recognition result corresponding to each part of the speech information is displayed. This is because it becomes clear at a glance which portion of the transcription text corresponds, and it can be easily confirmed whether or not all of the voice information has been transcribed without omission.

  Referring to FIG. 1, a transcription text creation support system according to an exemplary embodiment of the present invention includes a synchronization information generation unit 101, a synchronization information storage unit 102, a confirmation correction unit 103, a display device 104, an input device 105, and a voice output device. 106 and an output device 107.

  The synchronization information storage unit 102 is configured by a randomly accessible storage device such as a magnetic disk device, and has a large number of entries 1025. One entry 1025 includes a time information field 1021, a voice information field 1022, a recognition result field 1023, and a transcription text field 1024. In one entry 1025, in the recognition result field 1023, the recognition of the division unit when the recognition result obtained when the speech information included in the material content 111 to be transcribed is subjected to speech recognition processing is divided into smaller units than sentences. The result is stored. In the voice information field 1022, binary data of the voice information that is the basis of the recognition result stored in the recognition result field 1023 is stored. In the time information field 1021, the voice information field 1022 is stored. The time information for identifying the existing audio information from a series of audio information included in the material content 111 is stored. The contents of these three fields 1021, 1022 and 1023 are automatically set mechanically. The remaining transcription text field 1024 is a field that the operator himself sets interactively.

  The synchronization information generation unit 101 inputs material content 111 including at least audio information, performs audio recognition processing on the audio information included in the material content 111, and divides the obtained audio recognition result into units smaller than sentences. For each division unit, one entry 1025 of the synchronization information storage unit 102 is allocated, the recognition result of the division unit is stored in the recognition result field 1023, and the original voice information corresponding to the recognition result is stored in the voice information field 1022 Is stored in the time information field 1021, and the transcription text storage field 1024 is in an initial state (generally blank, but may be left as per the recognition result). deep. As a division unit, for example, a word can be used.

  For example, it is assumed that a recognition result “eating fish” is obtained as a result of performing speech recognition on a series of speech information that exists between the playback times t1 to t3 in the material content 111. In this case, “fish”, which combines “fish (common noun)” and “wo (each particle)”, is combined with one word, “eat (single verb stem)” and “ta (auxiliary verb)”. “Eat” is set as another word, and one entry 1025-1, 1025-2 is assigned to each word. Then, the recognition result “fish” is stored in the recognition result field 1023 of the entry 1025-1, the audio information AAA corresponding to the recognition result “fish” is stored in the audio information field 1022, and the material content of the audio information AAA is stored. When the playback time from the beginning in 111 exists between t1 and t2, for example, the time t1 is stored in the time information field 1021, and the transcription text field 1024 is blank. Similarly, the recognition result field 1023 of the entry 1025-2 stores the recognition result “eat”, the speech information field 1022 stores the speech information BBB corresponding to the recognition result “eat”, and the material of the speech information BBB If the playback time from the beginning in the content 111 is t2 to t3, the start time t2 is stored in the time information field 1021, and the transcription text field 1024 is blank.

  The confirmation / correction unit 103 displays a confirmation / correction screen 1046 on the display device 104 constituted by a liquid crystal display or the like, and allows the operator to edit the transcription text using the voice recognition result through the confirmation / correction screen 1046. On the confirmation and correction screen 1046, a time information display unit 1041, a recognition result display unit 1042, a speech waveform display unit 1043, and a transcription text display unit 1044 are arranged in parallel in the horizontal direction (X-axis direction). Scroll buttons 1045-1 and 1045-2 are provided for scrolling the contents of 1041 to 1044. The scroll button 1045-1 is a button (rewind button) for scrolling from the present to the past time side, and the scroll button 1045-2 is a button (fast forward button) for scrolling from the present to the future time side.

  In the time information display unit 1041, the time information stored in the time information field 1021 of each entry 1025 of the synchronization information storage unit 102 is displayed in order of time. The recognition result display unit 1042 displays each item of the synchronization information storage unit 102. The recognition results stored in the recognition result field 1023 of the entry 1025 are displayed in chronological order, and the audio waveform display unit 1043 displays the audio information stored in the audio information field 1022 of each entry 1025 of the synchronization information storage unit 102. Simplified waveform images are displayed in chronological order, and in the transcribed text display unit 1044, the transcribed texts stored in the transcribed text field 1024 of each entry 1025 in the synchronization information storage unit 102 are displayed in chronological order.

  The time axes of the display information in these four display units 1041 to 1044 are synchronized. In other words, the entry 1025-1 of the synchronization information storage unit 102 will be described as an example. The time t1 stored in the time information field 1021 of the entry 1025-1 and the time information field 1021 of the entry 1025-2 immediately after that are stored. When the stored time t2 is displayed on the time information display unit 1041, the guide line 1047 drawn in parallel to the Y-axis direction from the display position of the time t1 and the display position of time t2 are drawn in parallel to the Y-axis direction. The recognition result “fish” stored in the recognition result field 1023 of the entry 1025-1 is displayed in the recognition result display portion 1042 sandwiched between the guideline 1048 and the voice sandwiched between the guidelines 1047 and 1048. Stored in the voice information field 1022 of the entry 1025-1 in the waveform display portion 1043. The waveform image of the voice information being displayed is displayed, and the transcription text stored in the transcription text field 1024 of the entry 1025-1 is displayed in the portion of the transcription text display section 1044 sandwiched between the guidelines 1047 and 1048. The When the scroll button 1045-1 or 1045-2 is operated, the display information of the four display units 1041 to 1044 is scrolled in synchronization.

  The confirmation / correction unit 103 reproduces the audio information of the designated division unit in response to an audio reproduction instruction designating the division unit on the confirmation / correction screen 1046 from the input device 105, and is configured by a speaker or the like. Output from the audio output unit 106. As a method for instructing audio reproduction for designating a division unit, a method of clicking a reproduced audio waveform portion or recognition result portion of the division unit with a mouse can be considered. As a result, the worker can confirm the original voice information by listening with his / her ear for each division unit.

  Further, the confirmation correction unit 103 responds to the input operation of the transcription text specifying the division unit on the confirmation correction screen 1046 from the input device 105, in the portion corresponding to the division unit of the transcription text display unit 1044. The inputted transcription text is displayed. For the input operation of the transcription text, in addition to the direct input from the keyboard constituting the input device 105, a method of copying the recognition result displayed on the recognition result display unit 1042 to the text display unit 1044 can be used. In this way, the transcription text edited on the transcription text display unit 1044 is written into the transcription text field 1024 of the corresponding entry in the synchronization information storage unit 102.

  The contents of the transcription text field 1024 of each entry edited and stored in the synchronization information storage unit 102 as described above are the time in the time information field 1021 in response to the transcription text output instruction from the input device 105. It is also possible to sequentially read from the synchronization information storage unit 102 and output from the output device 107 such as a printer.

  Next, examples of each part of the transcription text creation support system according to the present embodiment will be described. First, an embodiment of the synchronization information generation unit 101 will be described in detail with reference to the drawings.

  Referring to FIG. 2, an example of the synchronization information generation unit 101 includes a material input unit 1011 that inputs the material content 111, and a voice that performs voice recognition processing on the audio information of the material content 111 input by the material input unit 1011. When the recognition unit 1012 and the speech recognition unit 1012 generate the recognition result text in a certain word unit, time information from the head of the material content 111 of the speech information that is the basis of the recognition result text is acquired. Output the time information acquisition unit 1013 and the speech recognition unit 1012 to receive the recognition result text and the original voice information, receive the time information from the time information acquisition unit 1013, and receive the received time information, voice information, One entry 1025 having a recognition result and a writing text field set to blank is stored as synchronization information. And a recording section 1014 for recording the 102.

  FIG. 3 is a flowchart showing a processing flow of the synchronization information generation unit 101, and FIG. 4 is a diagram showing a part of a data storage example of the synchronization information storage unit 102. Considering that the present invention is used for, for example, a transcription work for transcription of parliament video content, first, the parliament video as the material content 111 is reproduced by a video deck or the like and captured by the material input unit 1011. In general, the work of capturing video content into a computer is called capture. At the time of capture, the video content is digitized by MPEG1, which can be handled by a computer, and then temporarily stored in an external storage medium (not shown) such as a hard disk drive installed on the computer. (S101). At this time, reproduction time information may be added and recorded. Next, the voice recognition unit 1012 reads digital voice data from an external storage medium (not shown) in chronological order, performs voice recognition processing using an existing voice recognition technique, and generates a recognition result text (S102). Meanwhile, the time information acquisition unit 1013 receives the material content 111 from the material input unit 1011 in advance, associates the processing start time in the speech recognition unit 1012 with the time information of the material content 111, and determines which part of the current material content 111 Are being processed. Then, when the voice recognition unit 1012 recognizes the voice information and notifies the time information acquisition unit 1013 of the timing when the recognition result text is output, the time information acquisition unit 1013 matches the timing received from the voice recognition unit 1012. Time information from the top of the content 111 is acquired (S103). Here, the notification of the output timing may be made, for example, in units of words output by the voice recognition unit 1012.

  The voice recognition unit 1012 notifies the timing, and at the same time, notifies the recording unit 1014 of the recognition result text and the voice information that is the source thereof. The time information acquisition unit 1013 notifies the recording unit 1014 of the time information, and the recording unit 1014. Uses the received time information as a primary key, and stores the received voice information and recognition result text in association with the time information (S104). Further, the recording unit 1014 generates a field for storing the transcription text for each time information in the synchronization information storage unit 102.

  These series of steps are continuously performed up to the end of the material content 111.

  Referring to FIG. 4, next to the area where “0:00:03” to “0:00:14” are stored as time information, the voice information associated with the time information and the recognition result “ Continuing, “Autumn”, “At the Committee”, “Lending”, “Drafting”, “About”, “But”, and transcription text are stored. Here, the audio information 202 is illustrated as an image of an audio waveform for explanation.

  Next, an embodiment of the confirmation / correction unit 103 will be described in detail with reference to the drawings.

  Referring to FIG. 5, the confirmation / correction unit 103 receives the contents of each entry 1025 stored in the synchronization information storage unit 102, and receives the contents of each entry 1025 from the data reading unit 1031. The screen generation unit 1032 that generates the confirmation / correction screen 1046 as shown, and the audio information is received from the screen generation unit 1032, and the audio waveform that is visually easy to see for the operator is generated from the audio information in the binary format. The voice waveform generation unit 1033 for returning the waveform to the screen generation unit 1032, the display unit 1034 for displaying the confirmation correction screen 1046 generated by the screen generation unit 1032 on the display device 104, and the voice information from the data reading unit 1031, The audio reproduction unit 1035 and the audio reproduction unit 1035 that generate the audio signal that reproduces the audio signal The output unit 1036 for driving the audio output device 106 with the generated audio signal, the data writing unit 1037 for updating the contents of the transcription text field 1024 in each entry 1025 of the synchronization information storage unit 102, and a keyboard and a mouse. An input unit 1038 for inputting instructions and data from the input device 105 and a control unit 1039 for receiving instructions and data from the input unit 1038 and controlling each unit are provided.

  FIG. 6 is a diagram showing an example of the confirmation / correction screen 1046 in the initial state where the transcription text field 1024 of each entry in the synchronization information storage unit 104 is blank. Here, the time information display unit 1041, the recognition result display unit 1042, the speech waveform display unit 1043, and the transcription text display unit 1044 adopt a form in which they are arranged side by side in the application window. Using the time information stored in the information storage unit 102 as a key, each data is arranged on the same X-axis coordinate of the display.

  FIG. 7 is a flowchart showing a processing example of the confirmation / correction unit 103 when displaying the confirmation / correction screen in the initial state as shown in FIG. The screen generation unit 1032 of the confirmation correction unit 103 acquires time information from the head position of the material content 111 to be worked from the synchronization information storage unit 102 by the data reading unit 1031, and the window on the display device 104 by the display unit 1034. The time information is displayed on the time information display unit 1041 (S201). In FIG. 6, time information “0:00:03”, “0:00:04”,..., “0:00:14” are displayed in order.

  Next, the screen generation unit 1032 acquires audio information corresponding to the time information from the synchronization information storage unit 102 by the data reading unit 1031, and takes the format of digitized binary data by the audio waveform generation unit 1033. A voice waveform image is generated from the voice information using a general graphic equalizer display technique or the like, and the voice waveform is displayed on the voice waveform display unit 1043 in the window on the display device 104 by the display unit 1034 (S202). . Here, the audio waveform is displayed on the same X coordinate axis as the time information associated with the synchronization information storage unit 102 among the time information displayed on the time information display unit 1041. In FIG. 6, the waveform of the corresponding audio information is displayed on the same X coordinate axis as the time information “0:00:03”, “0:00:04”,..., “0:00:14”. .

  Next, the screen generation unit 1032 acquires the recognition result text corresponding to the time information from the synchronization information storage unit 102 by the data reading unit 1031, and the recognition result display unit 1042 in the window on the display device 104 by the display unit 1034. The recognition result text is displayed on (S203). Here, the recognition result text is displayed on the same X coordinate axis as the time information associated with the synchronization information storage unit 102 among the time information displayed on the time information display unit 1041. In FIG. 6, the recognition result text of the corresponding voice information is displayed on the same X coordinate axis as the time information “0:00:03”, “0:00:04”,..., “0:00:14”. ing.

  Finally, the screen generation unit 1032 acquires the transcription text corresponding to the time information from the synchronization information storage unit 102 by the data reading unit 1031, and the transcription text display unit in the window on the display device 104 by the display unit 1034. The transcription text is displayed in 1044 (S204). Here, the transcription text is displayed on the same X coordinate axis as the time information associated with the synchronization information storage unit 102 among the time information displayed on the time information display unit 1041. However, in the initial state, the transcription text information is not yet stored, so the transcription text display section 1044 corresponding to the time information is displayed as a blank.

  FIG. 8 is a flowchart showing the flow of processing executed by the confirmation / correction unit 103 after the flowchart shown in FIG. 7 is executed and the confirmation / correction screen 1046 in the initial state as shown in FIG. 6 is displayed on the display device 104. It is.

  First, the reproduction that the operator wants to hear by clicking the speech waveform of the speech waveform display unit 1043 displayed on the display device 104 or the recognition result text displayed on the recognition result display unit 1042 with the mouse of the input device 105. When a location is designated (YES in S301), the control unit 1039 that has detected this by a signal from the input unit 1038 acquires time information corresponding to the clicked speech waveform or recognition result text from the screen generation unit 1032. The audio reproduction unit 1035 notifies the audio reproduction unit 1035, the audio reproduction unit 1035 extracts the audio information corresponding to the notified time information from the synchronization information storage unit 102 through the data reading unit 1031 and generates a reproduction signal, and the output unit 1036 reproduces the reproduction information. The audio player 106 such as a speaker is driven by the signal (S302). Thereby, the audio | voice information corresponding to the location where the operator clicked is reproduced | regenerated as an audio | voice. Audio reproduction can be performed any number of times by repeating the same steps.

  The flow of the operation when the click operation on the speech waveform or the recognition result is not performed in step S301 will be described later (C).

  The operator listens to the audio information output from the audio output device 106 and determines whether the content of the recognition result text displayed on the recognition result display unit 1042 is correct. Here, the screen generation unit 1032 may have a function of modifying the recognition result text corresponding to the currently reproduced audio information by changing the color or changing the font. As a result, it is possible to visually confirm which part is currently being worked on, and thus it is possible to reduce visual work errors.

  When the recognition result is correct (YES in S303), the operator performs an operation of double-clicking the corresponding part of the recognition result display unit 1042 or copying the corresponding part to the text display unit 1044 by drag and drop. Then, the control unit 1039 detects this from the input signal of the input unit 1038 and controls the screen generation unit 1032 to copy and display the recognition result text in the blank space of the transcription text display unit 1044 (S304). . The state at this time is shown in FIG. Referring to FIG. 9, the recognition result “continue” corresponding to the time information “0:00:03” is clicked, and the result of confirming that the operator has heard the voice is normal. When dragging and dropping to the transcription text display portion 1044 on the same X coordinate axis, information of “continue” as a recognition result is additionally displayed on the transcription text display portion 1044. Thereafter, the transcription text additionally displayed on the transcription text display unit 1044 is stored in the synchronization information storage unit 103 in synchronization with time information as shown in FIG. 10 by the data writing unit 1037 (S306).

  On the other hand, when the recognition result is not correct (NO in S303), the operator inputs the correct text from the keyboard of the input device 105 to the corresponding portion of the transcription text display unit 1044 (S305). This is shown in FIG. Referring to FIG. 11, the recognition result “autumn” corresponding to the time information “0:00:04” is clicked, and the result of the operator confirming the voice is incorrect. The correct transcription text “first” is input to the display portion 1044 using the keyboard. Here, instead of a device that directly inputs characters like a keyboard, text may be input by speech recognition using a microphone. Thereafter, the transcription text additionally displayed on the transcription text display unit 1044 is stored in the synchronization information storage unit 103 by the data writing unit 1037 in synchronization with time information as shown in FIG. 12 (S306).

  If the work is completed up to the end of the work target content (YES in S307), the whole work is finished. If it is in the middle of the work target content (NO in S307), the work is continued (B). FIG. 13 is a flowchart illustrating an example of a process for continuing the transcription.

  Referring to FIG. 13, the operator clicks the position of the speech waveform display unit 1043, the recognition result display unit 1042, or the transcription text display unit 1044 displayed on the display device 104 with the mouse, and selects a desired reproduction part. When specified (YES in S401), the control unit 1039, which has detected this from the signal from the input unit 1038, acquires time information corresponding to the clicked speech waveform, recognition result text, or transcription text from the screen generation unit 1032. Then, the audio reproduction unit 1035 notifies the audio reproduction unit 1035, and the audio reproduction unit 1035 extracts the audio information corresponding to the notified time information from the synchronization information storage unit 102 to generate a reproduction signal, and the output unit 1036 uses the reproduction signal as a speaker. The audio player 106 is driven (S402). Thereby, the audio information corresponding to the location clicked by the worker is reproduced as audio. Audio reproduction can be performed any number of times by repeating the same steps.

  The flow of operation when there is no designation in step S401 will be described later (C).

  Next, the operator listens to the voice information output from the voice output device 106, and the content of the recognition result text displayed on the recognition result display unit 1042 and the transcription text displayed on the transcription text display unit 1044. Determine if the content of is correct. Here, the screen generation unit 1032 may have a function of modifying the recognition result text or the transcription text corresponding to the currently reproduced audio information by changing the color or changing the font. As a result, it is possible to visually confirm which part is currently being worked on, and thus it is possible to reduce visual work errors.

  If the transcribed text is blank or incorrect and the recognition result is correct (YES in S403), the operator double-clicks the corresponding part of the recognition result display part 1042, or transcribes the corresponding part by drag and drop. When an operation of copying to the text display unit 1044 is performed, the control unit 1039 detects this from the input signal of the input unit 1038 and controls the screen control unit 1032 to place the recognition result text in the place of the transcribed text display unit 1044. Copy and display (S404). On the other hand, if the transcription text is blank or incorrect and the recognition result is not correct (NO in S403), the operator puts the correct text on the corresponding portion of the transcription text display unit 1044 from the keyboard of the input device 105. Input (S405). Here, instead of a device that directly inputs characters like a keyboard, text may be input by speech recognition using a microphone.

  When the transcription text is inserted or inputted, the transcription text is notified to the data writing unit 1037 together with the time information corresponding to the transcription text, and the data writing unit 1037 accesses the corresponding entry in the synchronization information storage unit 102. Then, the received transcription text is stored in the synchronization information storage unit 102 in association with the received time information (S406).

  If it is in the middle of the work target content (NO in S407), the work is further continued (B).

  Referring to FIG. 14, “Continue”, “Previous”, “At the committee”, “Proposed”, “Proposal” are displayed on the transcript text display section 1044 on the same X coordinate axis corresponding to each time information. "About" and "But" information is reflected.

  If the work is completed up to the end of the work target content (YES in S407), the whole work is finished. As a method for notifying the end of work to the confirmation / correction unit 103, for example, as shown in FIG. 14, a work end button 1049 displayed as “Save and End” on the confirmation / correction screen 1046 is prepared. A method of notifying the end of the work to the confirmation and correction unit 103 when the operator presses a button can be used.

  FIG. 15 is a flowchart showing a flow of processing executed by the confirmation correction unit 103 when it is determined NO in step S301 of FIG. 8 and step S401 of FIG. Hereinafter, the scroll operation of the confirmation / correction screen 1046 will be described with reference to the flowchart of FIG. 15 and the example of the confirmation / correction screen of FIG. 16.

  If the worker presses the display icon 1045-1 or the display icon 1045-2 that are scroll buttons shown in FIG. 16 (YES in S501), the control unit 1039 detects this by an input signal from the input unit 1038. Then, the screen generation unit 1032 is notified of the change processing of the temporal display area for the scroll time set in advance. Here, for example, if the scroll time set when the display icon 1045-2 is pressed is 9 seconds, the screen generation unit 1032 acquires the start position of the time information currently displayed on the window, and the time The time information, voice information, recognition result, and transcription text are sequentially read out from the synchronization information storage unit 102 through the data reading unit 1031 so that the information plus 9th second is first (S502). The display is switched to data (S503).

  Referring to FIG. 14 and FIG. 16, in FIG. 14, the top time information of the time information display unit 1041 is “0:00:03”. Voice information corresponding to information “0:00:12”, recognition result “draft” “about” “but” “radish” “financial basis” “bad money” “accompanied”, transcription text “agenda” “About” “But” “Recent” “Financial foundation” “Deteriorating” “Accompanying” is displayed.

  The operation after the switching of the display shifts to the continued operation of the transcription shown in FIG. 13 (B). If there is no scroll instruction (NO in S502), it is confirmed whether or not the work is completed. If it is a continuous work (NO in S504), the process proceeds to the continuous work of transcription shown in FIG. 13 (B). If the work is finished (YES in S504), the work is finished.

  As described above, according to the present embodiment, it is possible to greatly reduce the burden on the operator when creating the transcription text of the audio information included in the material content 111, and to correct the accurate transcription text. Can be created.

  Although the embodiment of the present invention has been described above, the present invention is not limited to the above embodiment, and various other additions and modifications can be made. In addition, the synchronization information generation unit 101 and the confirmation correction unit 103 in the transcription text creation support system of the present invention can be realized by a computer and a program, as well as by hardware. The program is provided by being recorded on a computer-readable recording medium such as a magnetic disk or a semiconductor memory, and is read by the computer when the computer is started up, and the computer is controlled by controlling the operation of the computer. Function as the synchronization information generation unit 101 and the confirmation correction unit 103 in the embodiment.

It is a block diagram of an embodiment of a transcription text creation support system of the present invention. It is a block diagram of the Example of a synchronous information generation part. It is a flowchart which shows the flow of a process of a synchronous information generation part. It is a figure which shows some data storage examples of a synchronous information storage part. It is a block diagram of the Example of a confirmation correction part. It is a figure which shows an example of the confirmation correction screen in the initial state in which the transcription text field of each entry of the synchronous information storage part is blank. It is a flowchart which shows the process example of the confirmation correction part at the time of displaying the confirmation correction screen of an initial state. It is a flowchart which shows the flow of the process performed in a confirmation correction part, after the confirmation correction screen of an initial state is displayed on a display apparatus. It is a figure which shows the one scene where the transcription text was input into the confirmation correction screen. It is a figure which shows a mode that the transcription text was written in the synchronous information storage part. It is a figure which shows another scene where the transcription text was input into the confirmation correction screen. It is a figure which shows a mode that the transcription text was written in the synchronous information storage part. It is a flowchart which shows the process example of the continuous work of the transcription | transfer which a confirmation correction part performs. It is a figure which shows the other scene where the transcription text was input into the confirmation correction screen. It is a flowchart which shows the remaining process examples of a confirmation correction part. It is a figure which shows the example of the confirmation correction screen after scrolling.

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 ... Synchronization information generation part 1011 ... Material input part 1012 ... Voice recognition part 1013 ... Time information acquisition part 1014 ... Recording part 102 ... Synchronization information storage part 1021 ... Time information field 1022 ... Voice information field 1023 ... Recognition result field 1024 ... Writing Awakening text field 1025, 1025-1, 1025-2 ... entry 103 ... confirmation and correction unit 1031 ... data reading unit 1032 ... screen generation unit 1033 ... audio waveform generation unit 1034 ... display unit 1035 ... audio reproduction unit 1036 ... output unit 1037 ... Data writing unit 1038 ... input unit 1039 ... control unit 104 ... display device 1041 ... time information display unit 1042 ... recognition result display unit 1043 ... speech waveform display unit 1044 ... transcription text display unit 1045-1, 1045-2 ... scroll Tan 1046 ... confirmation correction window 1047,1048 ... guide lines 1049 ... working end button 105 ... input device 106 ... sound output unit 107 ... output device

Claims (7)

  Each of the storage device, the display device, the input device, the audio output device, and the voice recognition result obtained by performing the voice recognition processing on the voice information included in the material content is divided into smaller units than sentences. Synchronous information generation for storing the speech recognition result of the division unit, the original voice information, the time information for specifying the voice information, and the transcription text in the initial state in association with each division unit And the speech recognition result for each division unit stored in the storage device, the image of the original speech information, the time information for specifying the speech information, and the transcription text are aligned with each other on the time axis. And displays the audio information of the designated division unit in response to an audio reproduction instruction designating the division unit on the screen by the input device. Confirmation correction to update the transcription text of the designated division unit in the storage device in response to the input operation of the transcription text output from the voice output device and designating the division unit on the screen by the input device Transcription text creation support system characterized by comprising a part.   2. The transcription creation support system according to claim 1, wherein the division unit is a word.   The transcription text creation support system according to claim 1, wherein the speech information image is a speech waveform image generated from the speech information.   2. The transcription text creation support system according to claim 1, wherein the input operation of the transcription text designating the division unit on the screen by the input device is a copy operation for transcribing the recognition result to the text. .   In response to a scroll instruction from the input device, the confirmation / correction unit displays a voice recognition result for each division unit displayed on the screen of the display device, an image of the voice information that is the source, and the voice information. 2. The transcription text creation support system according to claim 1, wherein the time information for specifying and the transcription text are scrolled in synchronization with each other.   When a computer having a storage device, a display device, an input device, and an audio output device performs speech recognition processing on audio information included in material content, and divides the speech recognition result into smaller units than sentences For each division unit, the voice recognition result of the division unit, the original voice information, the time information for specifying the voice information, and the transcription text in the initial state are associated with each other and stored in the storage device. Information generating means, a speech recognition result for each division unit stored in the storage device, an image of the original speech information, time information for specifying the speech information, and a transcription text on the time axis Aligned and displayed in parallel on the screen of the display device, and in response to an audio reproduction instruction specifying the division unit on the screen by the input device, The voice information is reproduced and output from the voice output unit, and the designated division unit in the storage device is transcribed in response to the input operation of the transcription text designating the division unit on the screen by the input device. A program which functions as a confirmation and correction means for updating text.   The program according to claim 6, wherein the division unit is a word.

Images