JP6394332B2 - Information processing apparatus, transcription support method, and transcription support program - Google Patents

Description

  The present invention relates to an information processing apparatus, a transcription support method, and a transcription support program.

  In the transcription work to convert the recorded voice data into text, the transcription of voice is mainly manual, and even if automatic transcription is performed using voice recognition technology etc., the recognition accuracy is insufficient, Correction by manual transcription is required. In the case of manual transcription, it is necessary to listen again several times, and the cueing reproduction operation for that purpose is repeated, so that the transcription efficiency decreases. Therefore, there is a need for a technique for automating cue reproduction and improving transcription efficiency.

  For example, the speech position information in units of morphemes is listed from the speech recognition result of the speech data, and the cueing position of the speech data is specified based on the matching position by comparing the manually written text with the above-described list. There is a technique.

JP 2013-152365 A JP 2013-25763 A JP 2005-228178 A

  However, if there are many misrecognitions in the voice recognition result for the recorded voice data, the text does not match the comparison target text and the cue position cannot be specified.

  In one aspect, an object of the present invention is to appropriately specify a cueing reproduction position and improve document transcription efficiency.

  In one aspect, the information processing apparatus corresponds to the audio data storage unit that stores the audio data, the audio reproduction unit that reproduces the audio data from the cue reproduction position, and the audio data that is reproduced by the audio reproduction unit. A text generation unit that generates a transcription unit kana text from the input kana text, and a dynamic recognition dictionary based on one or more recognition vocabularies from the transcription unit kana text obtained from the text generation unit. A dictionary generation unit to generate, a position information extraction unit to extract position information of the voice data corresponding to the recognition vocabulary by voice recognition using the dynamic recognition dictionary obtained by the dictionary generation unit, and the position information A reproduction position determination unit that determines the cue reproduction position from the position information extracted by the extraction unit.

  As one aspect, the present invention can appropriately specify a cueing reproduction position, and can improve document transcription efficiency.

It is a figure which shows the function structural example of the information processing apparatus in 1st Embodiment. It is a figure which shows an example of a hardware constitutions. It is a flowchart which shows an example of the transcription assistance process in 1st Embodiment. It is a figure which shows the specific example of a transcription support process. It is a figure which shows the system configuration example of the information processing system in 2nd Embodiment. It is a figure which shows the system configuration example of the information processing system in 3rd Embodiment. It is a figure which shows 1st Example of a transcription assistance process. It is a figure for demonstrating the operation | movement in 1st Example of a transcription support process. It is a figure which shows 2nd Example of a transcription assistance process. It is a figure for demonstrating the operation | movement in 2nd Example of a transcription assistance process. It is a figure which shows 3rd Example of a transcription assistance process. It is a figure for demonstrating the operation | movement in 3rd Example of transcription support processing. It is a figure which shows 4th Example of a transcription assistance process. It is an example flowchart which shows 4th Example of a transcription assistance process. It is a figure for demonstrating the operation | movement in 4th Example of a transcription assistance process.

  Embodiments will be described below with reference to the drawings.

<First Embodiment>
FIG. 1 is a diagram illustrating a functional configuration example of the information processing apparatus according to the first embodiment. An information processing apparatus 10 illustrated in FIG. 1 includes a storage unit 11, a sound reproduction unit 12, an input unit 13, a text generation unit 14, a dictionary generation unit 15, a position information extraction unit 16, and a reproduction position determination unit 17. And have.

  The storage unit 11 stores various information necessary for performing the transcription support process, a processing result, and the like. The storage unit 11 includes, for example, an audio data storage unit 11a. The voice data storage unit 11a stores voice data that is a target for generating a transcription text or the like. The voice data is, for example, recorded voice or the like uttered by a person such as an interview, conversation, conference, lecture, speech, speech or the like, but is not limited thereto.

  The audio reproduction unit 12 reproduces audio data stored in the audio data storage unit 11a. The audio reproducing unit 12 may reproduce audio data at a normal speed, or may reproduce audio data in a low speed mode or a high speed mode. The playback speed can be changed before or during playback as specified by the user.

  The input unit 13 receives input of a transcription text or the like from a user or the like who has listened to the audio data reproduced by the audio reproduction unit 12. Further, the input unit 13 accepts inputs such as instructions for setting processing from the user to the information processing apparatus 10 and start / end processing of the transcription process. The input unit 13 is, for example, a keyboard, but is not limited thereto, and may be an operation button such as a touch panel displayed on the screen.

  The text generation unit 14 generates a kana text in a transcription unit for the information obtained from the input unit 13. The kana text of the transcription unit is, for example, text information input from the input unit 13 by the user's manual transcription after reproduction of recorded voice data, and is composed of, for example, one or a plurality of recognized vocabularies. The kana text is text information expressed in katakana, hiragana, etc. that has not been converted into kanji.

  The dictionary generation unit 15 generates a dynamic recognition dictionary having one or more recognition vocabularies from one or more transcript texts generated by the text generation unit 14. Note that the dictionary generation unit 15 may perform a long vowel generation process according to a preset long vowel generation rule and register the processed vocabulary (long phonetic kana text) in the dynamic recognition dictionary. The generated dynamic recognition dictionary may be stored in the storage unit 11 or the like, for example.

  The position information extraction unit 16 performs voice recognition processing using the dynamic recognition dictionary generated by the dictionary generation unit 15. Further, the position information extraction unit 16 extracts position information of the sound data corresponding to the recognized vocabulary from the sound data and the sound recognition result. In the speech recognition process, speech recognition of one recognition vocabulary is performed using a dynamic recognition dictionary. In the speech recognition processing, for example, it is preferable to perform word spotting in units of recognized vocabulary rather than dictation for recognizing speech data one by one. Word spotting is a method of picking up vocabulary in the dynamic recognition dictionary from the speech data, for example, and does not recognize extra vocabulary that is not in the dynamic recognition dictionary, so that erroneous recognition can be suppressed. Further, for example, an acoustic feature amount or the like can be used for the voice recognition process, but the present invention is not limited to this.

  The reproduction position determination unit 17 determines the cueing reproduction position of the audio data from the position information extracted by the position information extraction unit 16. The cue playback position may be the beginning or end of the fixed kana text, but is not limited to this, and may be the middle of the mora included in the kana text.

  The information processing apparatus 10 is a personal computer (PC), a server, a smartphone, a tablet terminal, or the like, but is not limited thereto.

  In the first embodiment, with the above-described configuration, for example, the playback position of the recorded voice following the confirmed kana text, the playback position of the recorded voice that is re-listen for confirmation, and the like can be appropriately specified, and the transcription efficiency of the document Can be increased. For example, in repeated re-listening in manual transcription or manual correction of automatic transcription, cue reproduction from the correct position becomes possible, and transcription efficiency is improved.

<Hardware configuration example>
Next, a hardware configuration example of a computer such as the information processing apparatus 10 will be described with reference to the drawings. FIG. 2 is a diagram illustrating an example of a hardware configuration. In the example of FIG. 2, the information processing apparatus 10 includes an input device 21, an output device 22, a drive device 23, an auxiliary storage device 24, a main storage device 25, a central processing unit (CPU) 26, and a network connection. Which are connected to each other by a system bus B.

  The input device 21 includes a keyboard and a pointing device such as a mouse operated by a user and a voice input device such as a microphone. The input device 21 activates a program execution instruction, various operation information, software, and the like from the user. Accept input of information.

  The output device 22 includes a display for displaying various windows, data, and the like necessary for operating the computer main body (information processing device 10) for performing the processing in the present embodiment. The output device 22 can display program execution progress, results, and the like by a control program of the CPU 26.

  Here, in the present embodiment, for example, the execution program installed in the computer main body such as the information processing apparatus 10 is provided by the recording medium 28 or the like. The recording medium 28 can be set in the drive device 23. Based on the control signal from the CPU 26, the execution program stored in the recording medium 28 is installed from the recording medium 28 into the auxiliary storage device 24 via the drive device 23.

  The auxiliary storage device 24 is, for example, a storage unit such as a hard disk drive (HDD) or a solid state drive (SSD). The auxiliary storage device 24 stores an execution program (for example, a transcription support program) in this embodiment, a control program provided in a computer, and the like based on a control signal from the CPU 26, and performs input / output as necessary. . The auxiliary storage device 24 can read and write necessary information from each stored information based on a control signal from the CPU 26 and the like.

  The main storage device 25 stores an execution program read from the auxiliary storage device 24 by the CPU 26. The main storage device 25 is a Read Only Memory (ROM), a Random Access Memory (RAM), or the like.

  The CPU 26 performs processing of the entire computer such as various operations and input / output of data with each hardware component based on a control program such as an operating system (OS) and an execution program stored in the main storage device 25. Control each process. Various information necessary during the execution of the program can be acquired from the auxiliary storage device 24, and the execution results and the like can also be stored.

  Specifically, the CPU 26 executes processing corresponding to the program on the main storage device 25 by executing the program installed in the auxiliary storage device 24 based on, for example, an instruction to execute the program obtained from the input device 21. Do.

  For example, the CPU 26 executes a transcription support program to store various types of information such as voice data in the storage unit 11 described above, voice reproduction by the voice reproduction unit 12, transcription text by the input unit 13, execution instructions, and the like. Input, text generation by the text generation unit 14, generation of a dictionary by the dictionary generation unit 15, extraction of position information by the position information extraction unit 16, determination of the reproduction position of the audio data by the reproduction position determination unit 17, etc. . The processing content in the CPU 26 is not limited to the above-described content. The contents executed by the CPU 26 are stored in the auxiliary storage device 24 or the like as necessary.

  The network connection device 27 communicates with other external devices via a communication network such as the Internet or a local area network (LAN). The network connection device 27 acquires an execution program, software, setting information, and the like from an external device or the like by connecting to a communication network or the like based on a control signal from the CPU 26. The network connection device 27 may provide an execution result obtained by executing the program to an external device or the like, or may provide the execution program itself in the present embodiment to the external device or the like.

  The recording medium 28 is a computer-readable recording medium that stores an execution program and the like as described above. The recording medium 28 is, for example, a semiconductor memory such as a flash memory, or a portable recording medium such as a CD-ROM or DVD, but is not limited thereto.

  By installing an execution program (for example, a transcription support program) in the hardware configuration shown in FIG. 2, the hardware resource and the software can cooperate to realize the transcription support process in the present embodiment. it can.

<Transcription support processing>
Next, the transcription support process will be described with reference to a flowchart. FIG. 3 is a flowchart showing an example of the transcription support process in the first embodiment. In the example of FIG. 3, the input unit 13 of the information processing apparatus 10 reproduces, for example, voice data to be transcribed text stored in the voice data storage unit 11 a and receives text input for the reproduced voice data (S 01). . Next, the text generation unit 14 determines from the received information whether the transcribed text is before kana-kanji conversion (S02). In the process of S02, for example, if the input unit 13 is a keyboard or the like, the determination can be made based on whether or not a predetermined key such as a keyboard conversion button or a space key is pressed, but is not limited thereto. For example, before the keyboard conversion button or the space key is pressed, it is determined that the kana-kanji conversion is not performed. If the conversion button or the space key is pressed, it is determined that the kana-kanji conversion has been completed. In addition, for example, by determining whether or not kanji is included from the contents after the conversion button or space key is pressed (for example, character code), it is determined whether or not it is before kana-kanji conversion. Also good.

  In the process of S02, when it is before Kana-Kanji conversion (YES in S02), a transcription text of Kana input is acquired (S03). If it is after Kana-Kanji conversion (NO in S02), the converted Kana-Kanji text is acquired (S04). Here, the kana / kanji text means, for example, a transcription text manually input.

  Next, the text generator 14 determines whether or not the kana / kanji text has been confirmed (S05). In the process of S05, it is determined whether or not the kana kanji text has been acquired, and whether or not it has been determined is determined.

  If the kana / kanji text is not confirmed in the process of S05 (NO in S05), the process returns to the process of S01. In the process of S05, when the kana-kanji text is confirmed (YES in S05), the text generation unit 14 generates the kana text acquired in the transcription unit (determined unit) (S06).

  Next, the text generation unit 14 combines a plurality of consecutively generated kana texts (S07). In the processing of S07, for example, when the kana text of the transcription unit is equal to or less than the predetermined number of mora, it is combined with the kana text of the past transcription unit. In the process of S07, when the kana text of the transcription unit is equal to or less than the predetermined number of mora, it may be combined with the kana text of the past transcription unit so as to have the predetermined number of mora. Moreover, in the process of S07, the kana text of the transcription unit may be connected at any time.

  Next, the dictionary generation unit 15 generates a dynamic recognition dictionary using the kana text obtained in the process of S07 as one recognition vocabulary (S08). Note that the dynamic recognition dictionary can achieve high speed and high accuracy of recognition processing by using, for example, one recognition vocabulary as a unit. Further, the dynamic recognition dictionary obtained by the process of S08 may be stored in the storage unit 11.

  Next, the position information extraction unit 16 performs speech recognition (word spotting) of one recognized vocabulary using the dynamic recognition dictionary obtained in the process of S08 (S09). In the processing of S09, misrecognition can be suppressed by performing word spotting in units of recognized vocabulary instead of dictation for recognizing each character.

  Next, the position information extraction unit 16 acquires at least one or more of the top mora head, the middle mora head, and the end mora end of the word spotting as voice data position information from the recognition result obtained in S09 (S10). Here, the head mora head is, for example, position information immediately before the voice of “shi” is output when one recognized vocabulary is “sudo-so-andi”. Further, the head of the intermediate mora is position information immediately before the voice of “O” is output in the case of, for example, “Shudou-Toshi and”. Further, the terminal mora terminal is position information immediately after the sound of “bi” is output in the case of “sudo-that and”, for example. The position information to be acquired may be selectively acquired based on a preset mode or the like, or all position information may be acquired regardless of the mode.

  Next, the reproduction position determination unit 17 determines the audio data position information acquired in the process of S10 as a cue reproduction position (S11), and starts audio reproduction from the decided cue reproduction position (S12).

  Here, it is determined whether or not to end the process (S13). If the process is not ended (NO in S13), the process returns to S01. Further, when the process is terminated due to the termination operation from the user or when the voice data is terminated (NO in S13), the transcription support process is terminated.

  As described above, in the first embodiment, the information processing apparatus 10 acquires a kana text in a transcription unit, and generates a dynamic recognition dictionary that uses a kana text in a transcription unit or more as one recognition vocabulary. In addition, the information processing apparatus 10 performs word spotting based on one recognized vocabulary in speech recognition, and determines a cue playback position from the speech range that has been word spotted. In the first embodiment, the recognition accuracy can be improved and the accuracy of the cueing reproduction position can be improved by the word spotting voice recognition limited to a multi-mora and one recognition vocabulary.

  By the above-described processing, it is possible to improve the accuracy of determining the cueing reproduction position and increase the transcription efficiency at the time of manual transcription or manual correction of automatic transcription. Therefore, it is possible to improve the accuracy of automatic playback and stop of voice.

<Specific examples of transcription support processing>
Next, a specific example of the above-described transcription support process will be described. FIG. 4 is a diagram showing a specific example of the transcription support process. 4A and 4B respectively show “kana text transcription unit”, “one recognition vocabulary of dynamic recognition dictionary”, “cue playback position”, and “cue playback when transcription unit is determined”. An example is shown. In the example of FIGS. 4A and 4B, the transcription unit of the kana text is, for example, “Shudosokochi”, “and”, “hono”,... Each time is input, a dynamic recognition dictionary is generated.

  In the example of FIGS. 4A and 4B, for the convenience of explanation, each kana text such as “Sudosokochi”, “and”, “hono”,. Each dynamic recognition dictionary is left as it is. The dynamic recognition dictionary in the first embodiment is, for example, each record in FIGS. 4 (A) and 4 (B).

  FIG. 4A shows an example in which one recognized vocabulary is a kana text having a predetermined number of mora (5 mora) or more. By setting the minimum number of mora for one recognized vocabulary, the word spotting accuracy can be improved. Further, in the example of FIG. 4A, time information after reproduction of audio data is shown as position information when the word spotting end mora end is set as the cue playback position. By managing the position information as time information, for example, it is possible to reproduce audio data using the position information at the end of the transcription unit as the cue reproduction position.

  FIG. 4B shows an example in which one recognized vocabulary is connected kana text. Thereby, the number of mora of one recognition vocabulary can be expanded and word spotting accuracy can be improved. In the example of FIG. 4B, time information is shown when the head of the intermediate mora in word spotting is set as the cue playback position. This makes it possible to perform cue playback from an intermediate mora or the like as a transcription reconfirmation mode. The intermediate mora head indicates the head mora of the next vocabulary connected in the transcription unit in the connected kana text. As shown in FIG. 4B, by adopting the intermediate position in the word spotting, it is possible to improve the accuracy of the cue playback position in the transcription efficiency-oriented mode.

  Also, as shown in FIG. 4C, in order to improve the cue playback position accuracy, when the position of 900 ms from the head is set as the cue playback position, if the mora is the end mora, the next mora information ( Since there is no acoustic feature value), the detection accuracy of the end boundary tends to deteriorate. However, when there is an intermediate mora, there is information (acoustic feature amount) of the intermediate mora (next mora) when determining the cueing reproduction position, so that the detection accuracy of the boundary between mora can be improved. An appropriate position such as 880 ms from the beginning can be set as the cue playback position.

Second Embodiment
Next, a second embodiment will be described. In the second embodiment, a transcription support process similar to that in the first embodiment described above is realized by a system configuration using a server and a client terminal.

  FIG. 5 is a diagram illustrating a system configuration example of an information processing system according to the second embodiment. The information processing system shown in FIG. 5 includes a server 31 and a client terminal 32. The server 31 and the client terminal 32 are information that can transmit and receive data via a communication network 33 typified by the Internet or a LAN. It is connected. The communication network 33 may be wired or wireless, or a combination thereof. In addition, the number of the server 31 and the client terminal 32 is not limited to this, For example, the some client terminal 32 may be connected with the server 31. FIG. The server 31 and the client terminal 32 have the above-described hardware configuration of the information processing apparatus.

  The server 31 includes a communication control unit 41, a storage unit 42, a dictionary generation unit 43, and a position information extraction unit 44. The storage unit 42 includes an audio data storage unit 42a.

  The communication control unit 41 transmits and receives data and the like through communication control with the client terminal 32 and other external devices via the communication network 33. The audio data storage unit 42 a stores audio data obtained from each client terminal 32. The dictionary generation unit 43 generates a dynamic recognition dictionary from the kana text of the transcription unit obtained from the client terminal 32.

  The position information extraction unit 44 extracts the position information of the voice data using the dynamic recognition dictionary, and transmits the voice data position information to the client terminal 32 by the communication control unit 41.

  The server 31 may be a PC, for example, or may be a cloud server configured by cloud computing having one or more information processing apparatuses, but is not limited thereto.

  The client terminal 32 includes a communication control unit 51, an input unit 52, a text generation unit 53, a playback position determination unit 54, an audio playback unit 55, and a storage unit 56. An audio data storage unit 56a is provided.

  The communication control unit 51 transmits and receives data and the like through communication control with the server 31 and other external devices via the communication network 33. The input unit 52 accepts a user's manual input of text data, an instruction to start or end processing, and the like. The text generation unit 53 generates a text in a transcription unit. Here, the communication control unit 51 transmits the text of the transcription unit generated by the text generation unit 53 and the audio data obtained from the audio data storage unit 56 a to the server 31 via the communication network 33.

  The reproduction position determination unit 54 determines the cue reproduction position from the position information of the audio data corresponding to the text and audio data of the transcription unit transmitted to the server 31 by the communication control unit 51. The audio reproducing unit 55 reproduces the audio data from the cueing reproduction position obtained by the reproduction position determining unit 54.

  The audio data storage unit 56 a stores audio data for each client terminal 32. The stored audio data is transmitted from the communication control unit 51 to the server 31 via the communication network 33.

  Using the configurations of the server 31 and the client terminal 32 in the second embodiment described above, processing similar to the transcription support processing in the first embodiment described above can be performed. For example, among the processes shown in FIG. 3, for example, the processes of S08 to S10 are executed on the server 31 side, and other processes are executed on the client terminal 32 side. Data transmission / reception between the server 31 and the client terminal 32 is performed by the communication control units 41 and 51.

  According to the second embodiment, for example, a cloud service type transcription support system can be provided. Further, according to the second embodiment, the processing load on the client terminal 32 side can be reduced as compared with the first embodiment.

<Third Embodiment>
Next, a third embodiment will be described. In the third embodiment, a system configuration divided into a server and a client terminal is used as in the second embodiment, but a part of the configuration is changed. In the following description, the same reference numerals are given to the same components as those in the second embodiment, and a specific description thereof is omitted here.

  FIG. 6 is a diagram illustrating a system configuration example of an information processing system according to the third embodiment. The information processing system 30 ′ in the third embodiment includes a server 31 ′ and a client terminal 32 ′, and the server 31 ′ and the client terminal 32 ′ are connected with information that allows data transmission / reception via the communication network 33. Has been.

  Comparing the third embodiment and the second embodiment, the server 31 ′ does not have the audio data storage unit 42 a shown in the server 31. Further, the client terminal 32 ′ has a feature amount extraction unit 61 as compared with the client terminal 32.

  The feature amount extraction unit 61 analyzes the sound data in units of a predetermined frame length and extracts a corresponding acoustic feature amount. The acoustic feature amount is, for example, a feature amount such as Mel Frequency Cepstrum Coefficient (MFCC, Mel frequency cepstrum coefficient), but is not limited thereto. For example, power (sound volume) with respect to the input voice, differential mel frequency cepstrum coefficient (DMFCC, differential mel frequency cepstrum coefficient, etc.) can be used.

  The communication control unit 51 transmits the transcription unit kana text and the acoustic feature amount to the server 31 ′. Further, the communication control unit 51 receives the position information of the audio data from the server 31 ′.

  In the server 31 ′, the communication control unit 41 receives the transcriptional unit kana text and the acoustic feature amount from the client terminal 32 ′. Further, the communication control unit 41 transmits the position information of the audio data to the client terminal 32 ′.

  The dictionary generation unit 43 generates a dynamic recognition dictionary from the transcription unit kana text transmitted from the client terminal 32 ′. The position information extraction unit 44 recognizes a voice from an acoustic feature amount in units of a predetermined frame length transmitted from the client terminal 32 ′, and extracts voice data position information. The communication control unit 41 transmits the position information extracted by the position information extraction unit 44 to the client terminal 32 ′ via the communication network 33.

  The playback position determination unit 54 of the client terminal 32 ′ determines the cue playback position from the position information of the audio data obtained from the server 31 ′, as in the second embodiment. The audio reproducing unit 55 reproduces the audio data from the cueing reproduction position obtained by the reproduction position determining unit 54.

  According to the third embodiment, for example, a cloud service type transcription support system can be provided. Further, according to the second embodiment, the processing load on the client terminal 32 side can be reduced as compared with the first embodiment. In the second and third embodiments described above, so-called distributed speech recognition (DSR) processing can be realized. In distributed speech recognition, high load processing of speech recognition is performed on the server 31 side, and light load processing is performed on the client terminal 32 side. Further, in the third embodiment, not the voice data but the acoustic feature amount is transmitted to the server 31 ′, so that not only the processing load on the client terminal 32 side but also the network communication load can be reduced.

<Specific Examples of Dictionary Generation Units 15 and 43 and Playback Position Determination Units 17 and 54>
Next, a specific example of the transcription support process using the dictionary generation units 15 and 43 and the reproduction position determination units 17 and 54 in the first to third embodiments described above will be described with reference to the drawings.

<First embodiment>
FIG. 7 is a diagram showing a first embodiment of the transcription support process. FIG. 7A shows a configuration example of the dictionary generation units 15 and 43 in the first embodiment. In the first embodiment, the dictionary generation units 15 and 43 include a transcription unit combining unit 71 and a dynamic recognition dictionary generation unit 72.

  The transcription unit combining unit 71 combines the kana texts of the transcription units until the kana text of the transcription units reaches a predetermined number of mora. The dynamic recognition dictionary generating unit 72 generates a dynamic recognition dictionary from the kana texts of the transcription units combined by the transcription unit combining unit 71.

  FIG. 7B is a flowchart showing an example of dictionary generation processing in the first embodiment. In the example of FIG. 7B, when the transcription unit coupling unit 71 inputs the kana text (transcription unit) (S21), the number of mora (n) of the recognized vocabulary is counted (S22), and the number of mora ( It is determined whether n) is equal to or greater than a predetermined number of mora (threshold value) (S23).

  If the number is not equal to or greater than the predetermined number of moras (NO in S23), the transcription unit combining unit 71 combines the kana text corresponding to the recognition result that has just succeeded in the recognition processing to increase the number of mora of the vocabulary to be recognized. (S24). Recognition accuracy can be improved by increasing the number of mora. In addition, after the process of S24, the process returns to the process of S22.

  If the number of moras is equal to or greater than the predetermined number of moras (YES in S23), the dynamic recognition dictionary generation unit 72 generates a prolonged kana text (S25), generates a dynamic recognition dictionary (S26), and generates the dynamic recognition dictionary. A dynamic recognition dictionary is output (S27). In the process of S27, the generated dynamic recognition dictionary may be stored in the storage unit 11, 42 or the like.

  FIG. 7C shows an example of a transcription unit corresponding to the first embodiment. In the first embodiment, in principle, the combination of kana text is a transcription unit. For example, if the predetermined number of mora (threshold value) is 5 mora, the number of mora when “Saisei” is input is 4, so it is combined with the next “Ichino” and becomes “Saisei Ichino”. At this time, the number of mora becomes 7 and becomes 5 mora or more, so no further coupling is performed.

  In the first embodiment, as shown in FIG. 7C, the kana text may be combined in units of a predetermined number of mora. In this case, for example, if the predetermined number of mora is 5 mora, for example, if only “Saisei” is 4 mora, 1 mora is insufficient, and if combined “Saisei Ichino” is 7 mora, 2 mora. Over. Therefore, in order to set the number of mora to 5 mora, for example, “Seiichino” that is 5 mora before the end of the text may be extracted to generate a dynamic recognition dictionary.

  Further, in the first embodiment, as a kana text combining condition, as shown in FIG. 7C, all the transcription units in which the voice data position information is determined may be combined. For example, if the kana texts are “Atamashishi”, “Saisei”, “Ichino”, and the location information of each is fixed, it will be combined as “Atamashisaiseiichinoichi” Also good.

  FIG. 8 is a diagram for explaining the operation in the first embodiment of the transcription support process. In the example of FIG. 8, the operation content using the sentence “setting accuracy of the cue playback position is degraded” is shown as an example sentence. FIGS. 8A to 8C show how a dynamic recognition dictionary is generated and how sound is reproduced for a kana text (transcription unit) input in time series. Shows how to go.

  First, in the part (A) of FIG. 8, the kana text “Atamashishi” input for the voice reproduction “Atamashi ...” is greater than or equal to a predetermined number of mora (5 mora). A dynamic recognition dictionary is generated as it is. Further, in the example of FIG. 8A, since the voice data position information of “Atamashi” has been confirmed (speech recognition is successful), the recorded voice following the confirmed kana text is reproduced.

  Next, in the part (B) of FIG. 8, since the kana text “saisei” input for the audio reproduction “saisei ...” is not more than the predetermined number of moras (5 mora), An example is shown in which a dynamic recognition dictionary is generated when the combined kana text of the wake-up unit becomes 5 mora or more. In the first embodiment, at the time of determining the voice data position information, as shown in FIG. 8 (B), the vocabulary (long phonetic kana text) subjected to the long vowel processing according to the preset long vowel rule is also included. To register in the dynamic recognition dictionary. Accordingly, in the dynamic recognition dictionary, as shown in FIG. 8 (B), two pieces of data “Amata Seisei” and “Amata Seisei” are generated.

  Next, in the part of FIG. 8C, since the kana text “Ichino” input for the audio reproduction “Ichino ...” is not more than the predetermined number of mora (5 mora), An example is shown in which a dynamic recognition dictionary is generated at the time when it becomes 5 mora or more by combining with the kana text of the wakeup unit. In the case of FIG. 8C as well, the recognition vocabulary to which the long vowelization rule shown in FIG. 8B is applied is added to the dynamic recognition dictionary. Accordingly, in the dynamic recognition dictionary, as shown in FIG. 8C, two data of “saisei ichino” and “saisei ichino” are generated. Thereafter, the transcription process is performed until the example sentence is completed in the same procedure as described above.

  As described above, by using the dynamic recognition dictionary generated by the first embodiment, it is possible to appropriately specify the playback position of the recorded voice following the text determined by manual transcription by word spotting of the recorded voice. Can improve the efficiency of document transcription.

<Second embodiment>
FIG. 9 is a diagram showing a second embodiment of the transcription support process. FIG. 9A shows a configuration example of the dictionary generation units 15 and 43 in the second embodiment. In the second embodiment, the dictionary generation units 15 and 43 include a transcription unit combining unit 71 and a dynamic recognition dictionary generation unit 72 as in the first embodiment described above.

  In the second embodiment, a dynamic recognition dictionary is generated using position information of voice data. The position information of the voice data may be, for example, voice data position information only at the beginning and end of the kana text (transcription unit), or may be data position information of the start or end position of each mora of the kana text (transcription unit). .

  FIG. 9B is a flowchart showing a second embodiment of the dictionary generation process. In the example of FIG. 9B, the dictionary generators 15 and 43 input kana text (S31) and input voice data position information (S32). Next, the dictionary generators 15 and 43 determine whether or not to combine the kana text from the position information of the voice data (S33). In the process of S33, when the voice data position information for the kana text input in the process of S31 is indeterminate, it is determined to combine the kana text. If the position information for the kana text is fixed, it is determined that the combination is not performed.

  Note that the processing of S33 is not limited to this. For example, in the process of S33, when the interval between the position information of the beginning and end of the kana text (transcription unit) is less than a predetermined value from the audio data position information (the reproduction time of the audio data corresponding to the kana text is short) Case). Also, for example, for each mora of kana text (transcription unit), when the interval between the end position of the previous mora and the start position of the subsequent mora is below a predetermined value (when the position (interval) between mora is short) ) May be combined.

  When combining the kana texts (YES in S33), the dictionary generators 15 and 43 combine the kana texts in units of transcription (S34). If the kana text is not combined after the process of S34 or in the process of S33 (NO in S33), the dictionary generators 15 and 43 generate a prolonged kana text (S35) and generate a dynamic recognition dictionary. (S36), a dynamic recognition dictionary is output (S37).

  FIG. 9C shows an example (transcription unit) of audio data position information in the second embodiment. In the second embodiment, only the start position 300 ms and the end position 900 ms of the first mora sound in the transcription unit may be used as the sound data position information. In the second embodiment, whether or not the position information of the voice data corresponding to the input kana text has been confirmed, and whether or not the kana text should be combined using the voice data position information as shown in FIG. 9C. Can be determined appropriately.

  FIG. 10 is a diagram for explaining the operation in the second embodiment of the transcription support process. In the example of FIG. 10, as in the first embodiment, “setting accuracy of the cue playback position is degraded” is used as an example sentence. In the example of FIG. 10, as in FIG. 8, how the dynamic recognition dictionary is generated and how the voice is reproduced for the kana text (transcription unit) input in time series. It shows how to go.

  First, in the part of FIG. 10A, the kana text “Atamashishi” input for the voice reproduction “Atamashishi ...” is generated as a dynamic recognition dictionary. Further, in the example of FIG. 10A, since the voice data position information of “Atamashishi” is confirmed (speech recognition success), the recorded voice following the confirmed kana text is reproduced.

  Next, in the part of FIG. 10 (B-1), voice recognition fails for voice playback “saisei ...”, and voice data position information is indeterminate. Therefore, in such a case, as shown in FIG. 10B-2, the speech recognition process is executed again by combining with the kana text of the transcription unit before the speech data position information is confirmed. When the voice data position information is confirmed (speech recognition is successful), a dynamic recognition dictionary is generated (“Adamashi Saisei”, “Adamashi Saisei”). Thereafter, the transcription process is performed until the example sentence is completed in the same procedure as described above.

  As described above, in the second embodiment, a dynamic recognition dictionary can be efficiently generated by ignoring an indeterminate part in the middle and combining it with the previous speech.

<Third embodiment>
FIG. 11 is a diagram showing a third embodiment of the transcription support process. FIG. 11A shows a configuration example of the reproduction position determination units 17 and 54 in the third embodiment. In the third embodiment, the reproduction position determination units 17 and 54 include a mode designation unit 81 and a reproduction position selection unit 82.

  The mode designation unit 81 accepts designation of a playback mode by user designation or the like. The reproduction mode includes, for example, a confirmation mode for confirming the contents of the kana text in the determined transcription unit.

  The reproduction position selection unit 82 selects the reproduction start position of the audio data based on the mode designated by the mode designation unit 81. For example, in the case of “confirmation mode ON”, the playback position selection unit 82 starts playback from the beginning of the current transcription unit. For example, the mode is a mode in which the transcription text portion is viewed again to confirm whether there is a transcription mistake, but is not limited thereto.

  In the case of “confirmation mode OFF”, the playback position selection unit 82 performs cue playback from the end of the current transcription unit. For example, it is a mode in which cueing reproduction is performed from the next position following the confirmed kana text (transcription unit) and the transcription efficiency is prioritized. Note that the playback mode is not limited to the above-described example.

  FIG. 11B is a flowchart showing a third embodiment of the transcription support process. In the example of FIG. 11B, the reproduction position determination units 17 and 54 accept designation of a reproduction mode (S41), and then input audio data position information (S42).

  Next, the playback position determination units 17 and 54 determine whether or not the position information is indeterminate (S43). If the position information is indeterminate (YES in S43), the playback start position is set as indeterminate. Information to that effect is output (S44), and the process ends. In the process of S44, the reproduction start position is not output.

  In addition, when the position information is not indefinite (NO in S43), the reproduction position determination units 17 and 54 determine whether or not the confirmation mode is ON according to the reproduction mode (S45). When the confirmation mode is ON (YES in S45), the reproduction start position is positioned at the beginning of the audio data corresponding to the input kana text (transcription unit) (S46). If the playback mode is not ON (NO in S45), the playback start position is positioned at the end of the audio data corresponding to the input kana text (transcription unit) (S47). The reproduction position determination units 17 and 54 output the reproduction start position after the processing of S46 and S47 (S48).

  FIG. 12 is a diagram for explaining the operation in the third embodiment of the transcription support process. In the example of FIG. 12, as in the first and second embodiments described above, “determination of setting accuracy of the cue playback position” is used as an example sentence.

  In the example of FIG. 12, for a kana text (transcription unit) input in chronological order, how a dynamic recognition dictionary is generated and how sound is reproduced according to the playback mode. It shows.

  First, FIG. 12A shows an example in which the above-described confirmation mode is ON as an example of the playback mode. The kana text “Atamashishi” input for the voice reproduction “Atamashishi ...” is generated as a dynamic recognition dictionary. In the example of FIG. 12A, since the confirmation mode is ON, as shown in FIG. 12B, the recorded sound is reproduced from the head of the determined “Amatashi”. As a result, it is possible to efficiently check the written contents.

  Also, in the part of FIG. 12B, as an example of the playback mode, for example, when the above-described confirmation mode is OFF, the audio data position information is determined for the audio playback “Adamashi Saisei ...” ( When the speech recognition is successful, corresponding dynamic recognition dictionaries (“Adamashi Saisei”, “Adamashi Saisei-”) are generated. At this time, since the confirmation mode is OFF, as shown in FIG. 12C, cue playback is performed from the end position of the transcription unit. Thereby, the next transcription can be performed quickly.

  In the example of FIG. 12C, the mora number is adjusted from the predetermined number of mora or the like for the kana text “Ichinosete ...” input for the voice reproduction “Ichinosete... After the determination, dynamic recognition dictionaries (“saisei ichino” and “saisei ichino” are generated. Transcription processing is performed until the example sentence is completed in the same procedure as described above.

  As described above, in the third embodiment, it is possible to reproduce from a reproduction position suited to the user's purpose in accordance with the reproduction mode, so that the document transcription efficiency can be increased.

<Fourth embodiment>
FIG. 13 is a diagram showing a fourth embodiment of the transcription support process. In the fourth embodiment, an example in which the above-described first to third embodiments are partially combined is shown.

  In the example of FIG. 13, two dictionary generation units 15-1 and 15-2, a position information extraction unit 16, and a reproduction position determination unit 17 are included. The dictionary generation unit 15-1 includes a dynamic recognition dictionary generation unit 72-1. The dictionary generation unit 15-2 includes a transcription unit combining unit 71 and a dynamic recognition dictionary generation unit 72-2. The playback position determination unit 17 includes a mode specification unit 81 and a playback position selection unit 82. Since each configuration has been described in each of the above-described embodiments, a specific description thereof is omitted here.

  In the fourth embodiment, a plurality of dictionary generation units 15-1 and 15-2 generate dynamic recognition dictionaries (dictionaries A and B) under different conditions, and the generated dynamic recognition dictionary is used to generate voice data. Extract location information. The different conditions described above may be based on, for example, the number of mora, and are conditions based on confirmation, uncertainness, etc. of the voice data position information. The conditions for generating the dictionary shown in the first to third embodiments described above. However, the present invention is not limited to this.

  In the fourth embodiment, the mode designation unit 81 accepts the designation of the confirmation mode, selects the reproduction position of the audio data based on the accepted content, and outputs the reproduction start position. Next, the transcription support process corresponding to the configuration shown in FIG. 13 will be described using a flowchart.

  FIG. 14 is a flowchart of an example showing a fourth embodiment of the transcription support process. In the example of FIG. 14, upon receiving input of kana text (transcription unit) (S 51), the dictionary generation unit 15-1 generates a prolonged kana text (S 52) and creates a dynamic recognition dictionary (dictionary A). Generate (S53).

  Next, the position information extraction unit 16 performs voice recognition processing using the dictionary A (S54), and determines whether the recognition is successful (S55). When the recognition is successful (YES in S55), the position information extraction unit 16 extracts the voice data position information (S56). If the recognition is not successful (NO in S56), the position information extraction unit 16 makes the voice data position information uncertain (S57).

  Next, the dictionary generation unit 15-2 determines whether there is voice data position information (S58). If there is voice data position information (YES in S58), the kana text is transcribed and combined in units. (S59). In addition, after the process of S59 or when there is no voice data position information in S58 (NO in S58), the dictionary generation unit 15-2 generates a prolonged kana text (S60).

  Next, the dictionary generation unit 15-2 generates a dynamic recognition dictionary (dictionary B) (S61). Next, the position information extraction unit 16 performs voice recognition processing using the dictionary B (S62), and determines whether the recognition is successful (S63). When the recognition is successful (YES in S63), the position information extraction unit 16 extracts the voice data position information (S64). If the recognition is not successful (NO in S63), the voice data position information is determined indeterminate (S65).

  After the process of S64 or S65, it is determined whether or not the position information is uncertain (S66). If it is uncertain (YES in S66), the reproduction start position is uncertain (S67), and information to that effect is displayed. The process ends without outputting or outputting anything (S67).

  If the position information is not indefinite (NO in S66), the playback position determination unit 17 determines whether or not the confirmation mode designated by the user in the mode designation unit 81 is ON (S68).

  When the confirmation mode is ON (YES in S68), the reproduction position determination unit 17 positions the reproduction start position at the beginning of the audio data corresponding to the input kana text (transcription unit) (S69). If the playback mode is not ON (NO in S68), the playback start position is positioned at the end of the audio data corresponding to the input kana text (transcription unit) (S70). The reproduction position determination unit 17 outputs the reproduction start position after the processes of S69 and S70 (S71).

  FIG. 15 is a diagram for explaining the operation in the fourth embodiment of the transcription support process. In the example of FIG. 15, as in the first to third embodiments described above, “determination of setting accuracy of the cue playback position” is used as an example sentence.

  In the example of FIG. 15, for a kana text (transcription unit) input in chronological order, how a dynamic recognition dictionary is generated and how sound is reproduced according to the playback mode. It shows. In the fourth embodiment, when voice recognition fails, the dynamic recognition dictionary is regenerated under different conditions and recognized again.

  First, in the part of FIG. 15A, the kana text “Atama” input for the voice reproduction “Atamashi ...” is generated as a dynamic recognition dictionary. In the example of FIG. 15A, since the voice data position information of “Atama” is confirmed (speech recognition is successful), the recorded voice following the confirmed kana text is reproduced.

  Next, in the part of FIG. 15 (B-1), the voice recognition fails for the voice reproduction “Dashisaisei ...”, and the voice data position information is indeterminate. In this case, the dynamic recognition dictionary is regenerated under different conditions. In the part of FIG. 15 (B-2), “tamanashi” is generated as the dynamic recognition dictionary.

  Next, when the voice recognition process is executed again using the regenerated dynamic recognition dictionary and the voice data position information is determined (successful voice recognition), as shown in FIG. The process can be continuously executed for “Saisei ...”. Thereafter, the transcription process is performed until the example sentence is completed in the same procedure as described above.

  As described above, in the fourth embodiment, even if the speech recognition fails, the dynamic recognition dictionary used at that time can be regenerated and the speech recognition process can be executed again, so that the operation is not interrupted. A dynamic recognition dictionary can be generated efficiently. In the fourth embodiment, the two dictionary generation units 15-1 and 15-2 are used. However, the dynamic recognition dictionary may be regenerated repeatedly until the voice recognition is successful. In that case, you may have three or more dictionary production | generation parts.

  As described above, according to the present embodiment, it is possible to appropriately specify the cueing reproduction position, and it is possible to improve the document transcription efficiency. For example, it is possible to appropriately specify the playback position of the recorded voice following the confirmed kana text, the playback position of the recorded voice when listening again for confirmation, and the document transcription efficiency can be improved. For example, in the present embodiment, in order to appropriately obtain the repeated playback position of the recorded voice subsequent to the confirmed kana text when manually transcribing the recorded voice, one or more recognized vocabulary words are used as a unit from the kana text of the determined transcription unit. The generated dynamic recognition dictionary is generated, the recorded voice is spotted using the generated dynamic recognition dictionary, and the playback position of the recorded voice following the text determined by manual transcription is specified. Also, in this embodiment, when one recognition vocabulary subject to word spotting has a short number of mora, the word spotting accuracy cannot be increased. Therefore, when one recognition vocabulary is short, the recognition result of the word spotting subject including the last successful recognition result is included. Increase the number of mora in the vocabulary by a predetermined amount or more. Therefore, the accuracy increases as the number of mora increases. Further, if the position cannot be specified even if the number of mora for word spotting is increased, the immediately previous recognition result may be included.

  According to the present embodiment, for example, it can be applied to control of automatic reproduction of recorded sound, automatic stop, speech speed conversion function, etc. in manual transcription and manual transcription of an automatic transcription correction portion.

  Although the embodiments have been described in detail above, the invention is not limited to the specific embodiments, and various modifications and changes can be made within the scope described in the claims. Moreover, it is also possible to combine a part or all of each Example mentioned above.

In addition, the following additional remarks are disclosed regarding the above Example.
(Appendix 1)
An audio data storage unit for storing audio data;
An audio playback unit for playing back the audio data from the cue playback position;
A text generation unit that generates a kana text in a transcription unit from a kana text input corresponding to the audio data reproduced by the audio reproduction unit;
A dictionary generation unit that generates a dynamic recognition dictionary in units of one or more recognition vocabularies from kana text in a transcription unit obtained from the text generation unit;
A position information extraction unit that extracts position information of the voice data corresponding to the recognition vocabulary by voice recognition using the dynamic recognition dictionary obtained by the dictionary generation unit;
An information processing apparatus comprising: a reproduction position determination unit that determines the cue reproduction position from position information extracted by the position information extraction unit.
(Appendix 2)
The dictionary generation unit
The information processing according to claim 1, wherein, when the recognition vocabulary is not equal to or greater than a predetermined number of mora, the number of mora of the recognition vocabulary is increased by combining with the recognition vocabulary corresponding to the previous successful recognition result. apparatus.
(Appendix 3)
The dictionary generation unit
3. The information processing apparatus according to appendix 1 or 2, wherein when the speech recognition for the recognition vocabulary cannot be performed, the recognition vocabulary is combined.
(Appendix 4)
The dictionary generation unit
The dynamic recognition dictionary is generated using the kana text of the transcription unit obtained from the text generation unit and the position information of the voice data obtained by the position information extraction unit. 4. The information processing apparatus according to any one of 3.
(Appendix 5)
The dictionary generation unit
The information processing apparatus according to appendix 4, wherein when the position information of the voice data with respect to the recognized vocabulary is not fixed, the information is combined with the immediately preceding recognized vocabulary.
(Appendix 6)
A mode designation unit for accepting designation of a mode for determining the cue playback position;
Any one of Supplementary notes 1 to 5, further comprising: a reproduction position selection unit that selects the start or end position of the current transcription unit as a reproduction position in accordance with the confirmation mode designated by the mode designation unit. The information processing apparatus according to item.
(Appendix 7)
The dictionary generation unit
Supplementary notes 1 to 3, wherein when the speech recognition fails, the dynamic recognition dictionary used for the speech recognition is regenerated under different conditions, and the regenerated dynamic recognition dictionary is used for speech recognition again. 7. The information processing apparatus according to any one of 6.
(Appendix 8)
A feature amount extraction unit that extracts the acoustic feature amount by analyzing the audio data in a predetermined frame unit;
The position information extraction unit
Supplementary notes 1 to 3, wherein position information of the voice data corresponding to the recognition vocabulary is extracted by voice recognition using the acoustic feature quantity obtained by the feature quantity extraction unit and the dynamic recognition dictionary. The information processing apparatus according to any one of 7.
(Appendix 9)
Information processing device
Play audio data from the cue playback position,
Generate kana text in the transcription unit from the kana text input corresponding to the reproduced voice data,
Generating a dynamic recognition dictionary with one or more recognition vocabulary as a unit from the generated kana text of the transcription unit;
Extracting position information of the speech data corresponding to the recognition vocabulary by speech recognition using the generated dynamic recognition dictionary,
A transcription support method, wherein the cue playback position is determined from the extracted position information of the audio data.
(Appendix 10)
Play audio data from the cue playback position,
Generate kana text in the transcription unit from the kana text input corresponding to the reproduced voice data,
Generating a dynamic recognition dictionary with one or more recognition vocabulary as a unit from the generated kana text of the transcription unit;
Extracting position information of the speech data corresponding to the recognition vocabulary by speech recognition using the generated dynamic recognition dictionary,
A transcription support program for causing a computer to execute processing for determining the cue playback position from the extracted position information of the audio data.

DESCRIPTION OF SYMBOLS 10 Information processing apparatus 11, 42, 56 Memory | storage part 11a, 42a, 56a Audio | voice data storage part 12,55 Audio | voice reproduction | regeneration part 13,52 Input part 14,53 Text generation part 15,43 Dictionary generation part 16,44 Position information extraction part 17, 54 Playback position determination unit 21 Input device 22 Output device 23 Drive device 24 Auxiliary storage device 25 Main storage device 26 CPU
27 Network connection device 28 Recording medium 30 Information processing system 31 Server 32 Client terminal 41, 51 Communication control unit 61 Feature amount extraction unit 71 Transcription unit coupling unit 72 Dynamic recognition dictionary generation unit 81 Mode specification unit 82 Playback position selection unit

Claims (8)

An audio data storage unit for storing audio data;
An audio playback unit for playing back the audio data from the cue playback position;
A text generation unit that generates a kana text in a transcription unit from a kana text input corresponding to the audio data reproduced by the audio reproduction unit;
A dictionary generation unit that generates a dynamic recognition dictionary in units of one or more recognition vocabularies from kana text in a transcription unit obtained from the text generation unit;
A position information extraction unit that extracts position information of the voice data corresponding to the recognition vocabulary by voice recognition using the dynamic recognition dictionary obtained by the dictionary generation unit;
An information processing apparatus comprising: a reproduction position determination unit that determines the cue reproduction position from position information extracted by the position information extraction unit. The dictionary generation unit
2. The information according to claim 1, wherein when the recognized vocabulary is not equal to or greater than a predetermined number of mora, the number of mora of the recognized vocabulary is increased by combining with a recognized vocabulary corresponding to the immediately succeeded recognition result. Processing equipment. The dictionary generation unit
2. The dynamic recognition dictionary is generated by using the kana text of the transcription unit obtained from the text generation unit and the position information of the voice data obtained by the position information extraction unit. Or the information processing apparatus of 2. A mode designation unit for accepting designation of a mode for determining the cue playback position;
4. A playback position selection unit that selects a start or end position of a current transcription unit as a playback position in accordance with the confirmation mode specified by the mode specification unit. The information processing apparatus according to item 1. The dictionary generation unit
The dynamic recognition dictionary used for the speech recognition is regenerated under different conditions when the speech recognition fails, and the speech recognition is performed again using the regenerated dynamic recognition dictionary. 5. The information processing apparatus according to any one of items 4 to 4. A feature amount extraction unit that extracts the acoustic feature amount by analyzing the audio data in a predetermined frame unit;
The position information extraction unit
The position information of the voice data corresponding to the recognition vocabulary is extracted by voice recognition using the acoustic feature quantity obtained by the feature quantity extraction unit and the dynamic recognition dictionary. The information processing apparatus according to any one of 1 to 5. Information processing device
Play audio data from the cue playback position,
Generate kana text in the transcription unit from the kana text input corresponding to the reproduced voice data,
Generating a dynamic recognition dictionary with one or more recognition vocabulary as a unit from the generated kana text of the transcription unit;
Extracting position information of the speech data corresponding to the recognition vocabulary by speech recognition using the generated dynamic recognition dictionary,
A transcription support method, wherein the cue playback position is determined from the extracted position information of the audio data. Play audio data from the cue playback position,
Generate kana text in the transcription unit from the kana text input corresponding to the reproduced voice data,
Generating a dynamic recognition dictionary with one or more recognition vocabulary as a unit from the generated kana text of the transcription unit;
Extracting position information of the speech data corresponding to the recognition vocabulary by speech recognition using the generated dynamic recognition dictionary,
A transcription support program for causing a computer to execute processing for determining the cue playback position from the extracted position information of the audio data.

Images