JP2017203953A - Data processing device, data processing system, data processing method and data processing program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce the possibility of confidential information leaking to the outside when uttered speech data is transmitted to a speech recognition server and also recognize the speech and turn it into text with sufficient accuracy, as well as reduce the load of a speech recognition and text conversion process performed by an internal information processing terminal and secure a processing speed.SOLUTION: A data processing device 1 comprises: a speech data replacement unit 26 for replacing specific part speech data included in acquired speech data with replacement part speech data different from the specific part speech data, and outputting it as speech data for conversion; a communication unit 30 for transmitting the speech data for conversion to a speech recognition server, and receiving text data converted from the speech data for conversion from the speech recognition server; and a text data reverse replacement unit 31 for extracting post-replacement text data, among the text data inputted from the speech recognition server, that corresponds to the replacement part speech data, and replacing the post-replacement speech data with pre-replacement text data that corresponds to the specific part speech data and outputting the resulting speech data.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a data processing apparatus, a data processing system, a data processing method, and a data processing program for converting audio data into text.

  In organizations such as corporations, when creating minutes of meetings, the contents of the speaker's statements are usually assigned by assigning one or more clerks and manually transcribing the contents heard by the clerk. It has been. However, due to the problem that this method is costly and lacks accuracy, a system has been proposed in which a minutes is automatically created using a speech recognition device (see Patent Document 1). In this technology, voice recognition processing is performed at an information processing terminal owned by each conference participant.

JP 11-272663 A

  By the way, in order to recognize and text the entire spoken language such as the minutes of a meeting with sufficient accuracy, high calculation capability such as an external server and sufficiently accumulated learning data stored therein are required. Therefore, in order to speech-recognize and text conference minutes with sufficient accuracy, it is necessary to transmit speech data of conference participants to an external server for speech recognition and text. However, when confidential information is included in the utterance content, there is a possibility that confidential information may be leaked to the outside if the utterance voice data is transmitted to an external voice recognition server.

  An object of the present invention is to reduce the possibility of leakage of confidential information to the outside when transmitting utterance voice data to a voice recognition server, and to enable voice recognition / text conversion with sufficient accuracy. It is to provide a data processing device, a data processing system, a data processing method, and a data processing program capable of reducing the processing load of voice recognition and text processing performed by the information processing terminal and ensuring the processing speed. .

The data processing apparatus of the present disclosure
A data processing device that outputs a recognition result of collected sound data,
An audio data replacement unit that replaces the specific unit audio data included in the collected audio data with replacement unit audio data different from the specific unit audio data and outputs the converted audio data;
A communication unit that transmits the voice data for conversion to a voice recognition server and receives text data converted from the voice data for conversion from the voice recognition server;
Of the text data input from the speech recognition server, extract post-replacement text data corresponding to the replacement unit speech data, and replace the post-replacement text data with pre-replacement text data corresponding to the specific unit speech data A text data reverse replacement unit that outputs as a recognition result of the collected sound data;
Is provided.

The data processing system of the present disclosure includes:
A data processing system that outputs a recognition result of collected sound data,
An audio data replacement unit that replaces the specific unit audio data included in the collected audio data with replacement unit audio data different from the specific unit audio data and outputs the converted audio data;
A communication unit that transmits the voice data for conversion to a voice recognition server and receives text data converted from the voice data for conversion from the voice recognition server;
Of the text data input from the speech recognition server, extract post-replacement text data corresponding to the replacement unit speech data, and replace the post-replacement text data with pre-replacement text data corresponding to the specific unit speech data A text data reverse replacement unit that outputs as a recognition result of the collected sound data;
Is provided.

The data processing method of the present disclosure includes:
A data processing method for outputting a recognition result of collected sound data,
A voice data replacement step of replacing the specific part voice data included in the collected voice data with replacement voice data different from the specific part voice data and outputting the voice data for conversion;
A communication step of transmitting the voice data for conversion to a voice recognition server and receiving text data converted from the voice data for conversion from the voice recognition server;
Of the text data input from the speech recognition server, extract post-replacement text data corresponding to the replacement unit speech data, and replace the post-replacement text data with pre-replacement text data corresponding to the specific unit speech data A text data reverse replacement step for outputting the collected sound data as a recognition result;
Is provided.

The data processing program of the present disclosure
A data processing program executed in a data processing device that outputs a recognition result of collected sound data,
For the computer of the data processing device,
A process of replacing the specific unit audio data included in the collected audio data with replacement unit audio data different from the specific unit audio data and outputting the converted audio data as conversion audio data;
Processing for transmitting the voice data for conversion to a voice recognition server and receiving text data converted from the voice data for conversion from the voice recognition server;
Of the text data input from the speech recognition server, the post-replacement text data corresponding to the replacement unit speech data is extracted, and the post-replacement text data is replaced with the pre-replacement text data corresponding to the specific unit speech data. Processing to output the collected sound data as a recognition result;
Is executed.

  According to the present invention, it is possible to reduce the possibility of leakage of confidential information to the outside when transmitting utterance voice data to the voice recognition server, and to perform voice recognition / text conversion with sufficient accuracy. In addition, the present disclosure can reduce the load of speech recognition / text processing performed by the internal information processing terminal, and can ensure the processing speed.

The figure which shows an example of the image of the place where the audio | voice processing system of this 1st Embodiment was installed The block diagram which shows the system configuration | structure of the speech processing system of this 1st Embodiment. The flowchart explaining the operation | movement procedure regarding the audio | voice data replacement of the audio | voice processing system of this 1st Embodiment. The flowchart explaining an example of the operation | movement procedure regarding the speech recognition server of the speech processing system of this 1st Embodiment. The flowchart which shows the operation | movement procedure regarding the text data reverse substitution of the speech processing system of this 1st Embodiment. The figure which shows the example of the audio | voice input process which the audio | voice input process part implements in this 1st Embodiment. The figure which shows the example of the speech area detection process which the speech area detection part implements in this 1st Embodiment. The figure which shows the example of the combination table | surface in this 1st Embodiment. The figure which shows the process example of the specific audio | voice data detection process in this 1st Embodiment. The figure which shows the example of the specific information utterance timetable in this 1st Embodiment The figure which shows the example of the substituted audio | voice data table in this 1st Embodiment, and the process example of a substituted audio | voice synthesis process The figure which shows the process example of the audio | voice data shift process in this 1st Embodiment. The figure which shows the process example of the audio | voice data replacement process in this 1st Embodiment. The figure which shows the example of the replacement history table | surface in this 1st Embodiment. The figure which shows the process example of the audio | voice data transmission process for a conversion in this 1st Embodiment, the audio | voice data reception process for a conversion, an audio | voice data conversion process, and a text data transmission process. The figure which shows the process example of the text data reverse substitution process in this 1st Embodiment, and a recognition result output process The figure which shows the example of the substituted audio | voice data table in this 2nd Embodiment, and a substituted audio | voice synthesis process. The figure which shows the process example of the audio | voice data shift process in this 2nd Embodiment. The figure which shows the process example of the audio | voice data replacement process in this 2nd Embodiment. The figure which shows the example of the substituted audio | voice data table and the substituted audio | voice synthetic | combination process in this 3rd Embodiment. The figure which shows the process example of the audio | voice data shift process in this 3rd Embodiment. The figure which shows the process example of the audio | voice data replacement process in this 3rd Embodiment.

  Hereinafter, an embodiment (hereinafter referred to as “the present embodiment”) that specifically shows a speech processing system according to the present invention will be described with reference to the drawings.

(First embodiment)
FIG. 1 is a diagram illustrating an example of an image of a place where the voice processing system according to the first embodiment is installed. FIG. 2 is a block diagram showing a system configuration of the voice processing system 1 according to the first embodiment. FIG. 3 is a flowchart for explaining an operation procedure related to voice data replacement in the voice processing system 1 of the first embodiment. FIG. 4 is a flowchart illustrating an example of an operation procedure related to the speech recognition server 10 of the speech processing system 1 according to the first embodiment. FIG. 5 is a flowchart showing an operation procedure related to reverse text data replacement in the speech processing system 1.

  The speech processing system 1 shown in FIGS. 1 to 5 is a speech input processing unit 3 (an omnidirectional microphone, a directional microphone) installed in a place 2 (for example, a conference room, a bank counter, or an office) where speech recognition is performed. , A headset, etc.) is picked up and the recognition result is output to the display unit 4. Of the two conference participants, one becomes the voice recognition speaker 5, and the voice input processing unit 3 collects the voice 6 of the voice recognition speaker 5. The utterance voice 6 is replaced by the conversion voice data 8 by the data processing device 7 and transmitted to the voice recognition server 10 via the network 9. The network 9 may be a wired network (for example, an intranet or the Internet), or may be a wireless network (for example, a wireless local area network (LAN)). The speech recognition server 10 transmits text data 11 corresponding to the conversion speech data 8 to the data processing device 7. The data processing device 7 reversely replaces the text data 11 received from the speech recognition server 10 with text data corresponding to the uttered speech 6, and displays the reversely replaced text data as a recognition result of the uttered speech 6. Output to. Further, the data processing device 7 may be provided with an operation unit 12 for performing an operation.

  The voice input processing unit 3 picks up the uttered voice 6 uttered by the voice recognition speaker 5 and outputs it as the collected voice data 13 to the utterance section detecting unit 14 (voice input processing S1).

  The utterance section detection unit 14 outputs the utterance section voice data 15 obtained by removing the noise portion before and after the utterance section of the voice recognition speaker 5 from the input collected voice data 13 to the specific voice data detection unit 16 (speaking section). Detection process S2).

  The specific voice data detection unit 16 generates uttered specific part voice data 20 based on the input utterance section voice data 15 and the combination table 19 stored in the combination storage unit 18 to detect the utterance time. It outputs to the part 17 (specific audio | voice data detection process S3). The utterance time detection unit 17 generates a specific information utterance time table 21 in which the utterance start time and the utterance end time of the specific unit audio data 20 are described, and outputs the specific information utterance time table 21 to the replacement voice synthesis unit 22. (Speech time detection process S4).

  The replacement speech synthesizer 22 generates a replacement speech data table 24 including the replacement portion speech data 23 based on the input specific information utterance time table 21 and outputs it to the speech data shift portion 25 (replacement speech synthesis processing S5). .

  The voice data shift unit 25 generates the shift utterance section voice data 27 based on the input utterance section voice data 15 and the replacement voice data table 24, and outputs it to the voice data substitution section 26 (voice data shift processing S6). . By the voice data shift process S6, the conversion voice data 8 including the voice data after the replacement is generated in a natural manner as before the replacement, so that the accuracy of voice recognition / text conversion can be ensured.

  The voice data replacement unit 26 stores a replacement history table 28 generated based on the input shift utterance interval voice data 27 and the replacement voice data table 24 in the replacement history storage unit 29, and converts the conversion voice data 8 into the communication unit. 30 (audio data replacement processing S7). Providing the replacement history storage unit 29 facilitates processing when creating pre-replacement text data from text data.

  The communication unit 30 transmits the input conversion voice data 8 to the voice recognition server 10 via the network 9 (conversion voice data transmission process S8).

  The voice recognition server 10 receives the conversion voice data 8 via the network 9 (conversion voice data reception process S11), and converts the received conversion voice data 8 into text data 11 (voice data conversion process S12). The converted text data 11 is transmitted to the network 9 (text data transmission process S13). The communication unit 30 of the data processing device 7 outputs the text data 11 received via the network 9 to the text data reverse replacement unit 31 (text data reception processing S21).

  The text data reverse replacement unit 31 generates a recognition result to be originally obtained based on the input text data 11 and the replacement history table 28 stored in the replacement history storage unit 29 (text data reverse replacement processing). S22), the recognition result is output to the display unit 4 (recognition result output process S23). If the operation unit 12 is provided, the user edits the combination table 19 stored in the combination storage unit 18 via the operation unit 12 so that the specific voice to be replaced is appropriately designated by text and replaced. Later words can be specified easily. The operation unit 12 is, for example, a mouse or a keyboard.

  FIG. 6 is an example of the voice input process S1 performed by the voice input processing unit 3 in the first embodiment. For example, the voice input processing unit 3 includes a device capable of collecting sound such as an omnidirectional microphone, a directional microphone, and a headset, and collects ambient environmental noise sound and speech sound 6 of the speech recognition speaker 5. The voice input processing unit 3 converts the collected voice into an analog electric signal, further converts the analog electric signal into digital voice data (pulse code modulation (PCM) or the like), and outputs the collected voice data 13.

  FIG. 7 is an example of the speech segment detection process S2 performed by the speech segment detection unit 14 in the first embodiment. The collected voice data 13 output from the voice input processing unit 3 is input to the utterance section detection unit 14. The voice data before and after the time zone in which the voice recognition speaker 5 is speaking includes only the environmental noise voice. Since it is not necessary to perform speech recognition for the environmental noise speech part, the speech segment detection unit 14 determines that the speech recognition speaker 5 of the collected speech data 13 is based on the characteristics (waveform amplitude, frequency band, etc.) of the collected speech data 13. The voice data of the uttered section is cut out as the utterance section voice data 15 and output.

  FIG. 8 is a diagram illustrating an example of the combination table 19 in the first embodiment. The combination table 19 includes a combination ID column, a text data string before replacement, and a text data string after replacement. In the combination ID column, an identification number for uniquely identifying the combination is described. In the pre-replacement text data string, a short word or phrase detected by the specific voice data detection unit 16 in order to detect the specific part voice data 20 in the utterance section voice data 15 is described. In the post-replacement text data string, information for creating a replacement speech data table 24 that the replacement speech synthesizer 22 refers to in order to generate the replacement portion speech data 23 is described.

  In all combinations described in the combination table 19, when the voice data replacing unit 26 outputs the converted voice data 8 from the collected voice data 13 in the pre-replacement text data string and the post-substitution text data string, for conversion It is necessary to describe so that the context of the sentence in the audio data 8 does not become unnatural. For example, the sentence of the conversion voice data 8 in which “Fukuoka branch” is replaced with “Saudi Arabia branch” is natural, but the sentence of the conversion voice data 8 in which “Fukuoka branch” is replaced with “1000 branch” is unnatural. is there. When the context of the conversion voice data 8 is unnatural, some unnatural context may affect the entire text data 11 output by the voice recognition server 10 in the processing of the voice recognition server 10.

  For example, if “[Fukuoka] branch” is described in the pre-replacement text data string of the combination table 19 and “Saudi Arabia” is described in the post-replacement text data string, the description is in the utterance “Fukuoka branch”. This means that the corresponding specific part voice data 20 is replaced with the replacement part voice data 23 corresponding to the utterance “Saudi Arabia branch”. For example, when “sales is [*] yen” in the speech text data string before replacement and “random value (1, 10000)” is described in the text data string after replacement, the description is “ The specific part voice data 20 corresponding to utterances such as “100 million yen in sales” or “50 million yen in sales” is replaced with replacement part voice data 23 corresponding to utterances such as “sales is 1000 yen”. Means. Here, “[*]” means arbitrary utterance content, and “random value (1, 10000)” means that an integer between 1 and 10000 is randomly selected.

  FIG. 9 is a diagram illustrating a processing example of the specific voice data detection processing S3 in the first embodiment. In the specific voice data detection process S <b> 3, the utterance section voice data 15 including the specific part voice data 20 and the combination table 19 are input to the specific voice data detection unit 16 and the utterance time detection unit 17. The specific voice data detection unit 16 detects the specific part voice data 20 from the speech segment voice data 15. The utterance time detection unit 17 outputs the specific information utterance time table 21 including the calculated utterance start time and utterance end time, the combination ID of the combination table 19, the text data before replacement, and the text data after replacement.

FIG. 10 is a diagram illustrating an example of the specific information utterance time table 21 in the first embodiment.
The specific information utterance time table 21 includes a combination ID string, a pre-replacement text data string, a post-replacement text data string, an utterance start time string, and an utterance end time string.

  The specific information utterance timetable 21 shown in FIG. 10 is created on the basis of the utterance section voice data 15 corresponding to the utterance “The result of 2015, the sales of the Fukuoka branch resulted in 100 million yen”. is there. The combination ID “4” in the specific information utterance time table 21 corresponds to the combination ID “4” in the combination table 19 shown in FIG. In the “pre-replacement text data” column of the line in which the ID “4” of the specific information utterance time table 21 shown in FIG. 10 is described, “[Fukuoka]” described in the pre-replacement text data column of the combination table 19 is displayed. Instead of “branch”, only “Fukuoka” described in [] is described, and “Saudi Arabia” is described in the “Substitution text data” column. In the columns of “Speech start time” and “Speech end time” in the row where the combination ID “4” is written, the calculated “Fukuoka” utterance start time “2.0” and utterance end time are displayed. “2.3” is respectively described. In this way, a row in which the ID “4” of the specific information utterance time table 21 is described is generated.

  Similar to the combination ID “4”, the combination ID “1” of the specific information utterance time table 21 shown in FIG. 10 corresponds to the combination ID “1” of the combination table 19 shown in FIG. In the “pre-replacement text data” column of the line in which the ID “1” of the specific information utterance time table 21 shown in FIG. 10 is described, “sales is [ Instead of “*] Yen” (* = 100 million detected here), only “100 million” detected within [] is described. Then, in the column of “post-replacement text data” column of the specific information utterance time table 21 shown in FIG. 10, instead of “random value (1, 10000)” described in the pre-replacement text data column of the combination table 19 “1000” selected as a random value is described. In the column of “speech start time” and “speech end time” in the row in which the combination ID “1” is described, the calculated “100 million” utterance start time “3.5”, utterance end time “ 3.7 "is described respectively. In this way, a row in which the ID “1” of the specific information utterance time table 21 is described is generated.

  FIG. 11 is a diagram illustrating a processing example of the replacement speech data table 24 and the replacement speech synthesis processing S5 in the first embodiment. The replacement speech synthesizer 22 generates and outputs a replacement speech data table 24 based on the input specific information utterance time table 21. The generation of the replacement voice data table 24 will be described later. The generated replacement voice data table 24 includes a “replacement unit ID” column, a “replacement unit voice data content” column, an “utterance start time” column, an “utterance end time” column, a “replacement unit voice data length” column, “ It consists of a “shift amount” column and a “replacement unit audio data” column. In the “replacement unit ID” column, an identification number for uniquely identifying the replacement unit audio data 23 is described. In the “replacement unit audio data content” column, the utterance content of the replacement unit audio data 23 is described. In the “speech start time” column, a replacement start position when replacing the replacement unit voice data 23 with the shift utterance section voice data 27 is described. The “speech end time” column describes the replacement end position when replacing the replacement unit voice data 23 with the shift utterance section voice data 27. In the “replacement unit audio data length” column, the utterance length of the replacement unit audio data 23 is described. In the “shift amount” column, a numerical value indicating how much the audio data is shifted in the audio data shift processing S6 is described. The replacement unit audio data 23 is stored in the “replacement unit audio data” column.

Next, the generation of the replacement voice data table 24 will be described with some examples.
In FIG. 11, when the specific information utterance time table 21 is input to the replacement speech synthesizer 22, the replacement speech synthesizer 22 reads the first line of the input specific information utterance time table 21, that is, the combination ID “4”. Items relating to the pre-replacement text data and the post-replacement text data in the row are described in the first row of the replacement speech data table 24, that is, in each column in the replacement portion ID “1” row. “Saudi Arabia” in the “substitution text data” column in the combination ID “4” row of the specific information utterance time table 21 is “replacement portion voice data content” in the substitution portion ID “1” row of the substitution voice data table 24. Listed in the column. “2.0” in the “utterance start time” column of the combination ID “4” row of the specific information utterance time table 21 is the “utterance start time” column in the replacement part ID “1” row of the replacement voice data table 24. It is described in. “2.3” in the “utterance end time” column of the combination ID “4” row of the specific information utterance time table 21 is the “utterance end time” column in the replacement part ID “1” row of the replacement voice data table 24. It is described in. In the “replacement unit audio data” column of the replacement unit ID “1” row, the sound data of the speech uttering “Saudi Arabia” described in the “replacement unit audio data content” column at a natural speaking speed is displayed. Stored as audio data 23. Then, “0.5”, which is the length of the sound data of “Saudi Arabia”, is entered in the “replacement unit audio data length” column of the replacement unit ID “1” row. Further, in the “shift amount” column of the replacement unit ID “1” row, a predetermined formula, for example, a value “0.2” calculated from the replacement unit voice data length− (utterance end time−utterance start time) is entered. Is done.

  Similarly to the first line, the replacement speech synthesizer 22 selects items related to the pre-replacement text data and the post-replacement text data in the second line of the input specific information utterance time table 21, that is, the combination ID “1” line. Are described in the second row of the replacement voice data table 24, that is, in each column in the replacement portion ID “2” row. “1000” in the “replacement portion text data” column in the combination ID “1” row of the specific information utterance time table 21 is “replacement portion speech data content” in the replacement portion ID “2” row of the replacement speech data table 24. Listed in the column. However, the “utterance start time” column in the replacement part ID “2” row of the replacement voice data table 24 is “3” in the “utterance start time” column of the combination ID “1” row of the specific information utterance time table 21. .5 ”describes a value“ 3.7 ”obtained by adding the value“ 0.2 ”described in the“ shift amount ”column of the replacement unit ID“ 1 ”row. Similarly, in the “utterance end time” column in the replacement part ID “2” row of the replacement voice data table 24, the “utterance end time” column in the combination ID “1” row of the specific information utterance time table 21 is displayed. A value “3.9” obtained by adding the value “0.2” described in the “shift amount” column of the replacement unit ID “1” row is described in “3.7”. In the “replacement unit audio data” column of the replacement unit ID “2” row, the sound data of the speech that speaks “1000” described in the “replacement unit audio data content” column at a natural speaking speed is replaced with the replacement unit. Stored as audio data 23. Then, “0.1”, which is the length of the sound data of “1000”, is entered in the “replacement unit audio data length” column of the replacement unit ID “2” row. Further, in the “shift amount” column in the replacement unit ID “2” row, a value “−0.1” calculated from a predetermined formula, for example, replacement unit voice data length− (utterance end time−utterance start time) is set. Filled in.

  FIG. 12 is a diagram illustrating a processing example of the audio data shift processing S6 in the first embodiment. The voice data shift unit 25 generates and outputs shift utterance section voice data 27 based on the input utterance section voice data 15 and the replacement voice data table 24. Since “2.3” is described in the utterance end time of the first row of the replacement voice data table 24 input to the voice data shift unit 25, “0.2” is described in the shift amount column. The voice data shift unit 25 shifts all the voice data after 2.3 seconds of the input speech section voice data 15 in the positive direction (that is, backward) by 0.2 seconds. Further, “3.9” is described in the utterance end time of the second row of the replacement audio data table 24 input to the audio data shift unit 25, and “−0.1” is described in the shift amount column. Therefore, the voice data shift unit 25 shifts all the voice data after 3.9 seconds of the input speech section voice data 15 in the negative direction (that is, before) by 0.1 seconds. In this way, the voice data shift unit 25 outputs the voice data after the voice data shift process S6 is completed as the shift utterance section voice data 27.

  As a result, the replacement voice data 8 including the voice data after the replacement is generated in a natural manner as before the replacement, so that the accuracy of voice recognition / text conversion can be ensured.

  FIG. 13 is a diagram illustrating a processing example of the audio data replacement processing S7 in the first embodiment. The voice data replacement unit 26 generates and outputs a replacement history table 28 and conversion voice data 8 based on the input shift utterance section voice data 27 and the replacement voice data table 24. Since “2.0” is described in the utterance start time column of the first row of the replacement speech data table and “0.5” is described in the replacement portion speech data length column, the shift speech segment speech data 27 includes: The sound data of 2.0 seconds to 2.5 seconds is replaced with the sound data of the replacement unit sound data 23 stored in the replacement unit sound data string in the first row of the replacement sound data table 24. The same process is performed for the second row of the replacement voice data table 24, and the replaced voice data is output as the conversion voice data 8.

  FIG. 14 is a diagram showing an example of the replacement history table 28 in the first embodiment. The replacement history table 28 includes “before replacement” columns and “after replacement” columns. For example, “Fukuoka” is described in the “before replacement” column of the first row of the replacement history table 28, “Saudi Arabia” is described in the “after replacement” column, and the “before replacement” column of the second row of the replacement history table 28 is displayed. When “100 million” is described in the “after replacement” column, “1000” is described, the replacement part voice corresponding to “Saudi Arabia” is the specific part voice data 20 corresponding to “Fukuoka” in the shift utterance section voice data 27. The conversion voice data 8 is generated in which the specific part voice data 20 corresponding to 27 “100 million” of the shift utterance section voice data is replaced with the replacement part voice data 23 corresponding to “1000”. Means that.

  FIG. 15 is a diagram illustrating processing examples of the conversion voice data transmission process S8, the conversion voice data reception process S11, the voice data conversion process S12, and the text data transmission process S13 in the first embodiment. In FIG. 15, the communication unit 30 transmits the input conversion voice data 8 to the outside of the data processing device 7 (for example, the voice recognition server 10). Upon receiving the conversion voice data 8, the voice recognition server 10 converts the conversion voice data 8 into corresponding text data 11 and outputs it to the communication unit 30. For example, when the conversion voice data 8 is input, such as “results in 2015, sales of Saudi Arabia branch is 1000 yen”, “results in 2015, sales of Saudi Arabia branch is 1000 yen. The converted text data 11 is output to the communication unit 30. The communication unit 30 transmits the input text data 11 to the data processing device 7. Examples of the transmission / reception path between the communication unit 30 and the voice recognition server 10 include Ethernet, USB, RC232 (serial communication), and the like.

  FIG. 16 is a diagram illustrating a processing example of text data reverse replacement processing S22 and recognition result output processing S23 in the first embodiment. The text data reverse replacement unit 31 outputs a recognition result 32 to be originally obtained based on the input text data 11 and the replacement history table 28. The input text data 11 is “result of 2015, sales of Saudi Arabia branch was 1000 yen”, and “Fukuoka” in the “before replacement” column in the first row of the replacement history table 28 is “ When “Saudi Arabia” is described in the “After Replacement” column, “100 million” is described in the “Before Replacement” column of the second row, and “1000” is described in the “After Replacement” column, the text data reverse replacement unit 31 "Saudi Arabia" included in the text data 11 that was entered was replaced with "Fukuoka" and "1000" was reversely replaced with "100 million". "FY 2015 results, sales of the Fukuoka branch were 100 million yen. Is generated and output to the display unit 4. The display unit 4 is, for example, a display device or a printer device.

  Note that the user can create the combination table 19 according to the usage status by editing or newly creating the combination table 19 via the operation unit 12. As a result, the user can easily specify the specific part voice data 20 to be replaced by text and can easily specify the word after replacement.

  As described above, according to the first embodiment, the possibility of leakage of confidential information to the outside when transmitting utterance voice data to the voice recognition server is reduced, and voice recognition / text conversion is performed with sufficient accuracy. can do. In addition, it is possible to reduce the load of voice recognition / text processing performed by the internal information processing terminal, and to secure the processing speed.

(Second Embodiment)
Regarding the second embodiment, only the parts different from the first embodiment will be described. The other points are the same as in the first embodiment.

  FIG. 17 is a diagram illustrating an example of the replacement speech data table 24 and the replacement speech synthesis process S5 in the second embodiment. In FIG. 17, the second embodiment is different from the first embodiment in a calculation method of the replacement unit audio data length and a generation method of the audio data of the replacement unit audio data 23. The replacement speech synthesizer 22 generates and outputs a replacement data table based on the input specific information utterance time table 21. At this time, unlike FIGS. 11 and 12 of the first embodiment, the replacement speech data table For each of the 24 lines, a value obtained by subtracting the utterance start time from the utterance end time is described as the replacement unit voice data length. Specifically, the replacement speech synthesizer 22 selects “speech” from the value “2.3” in the “speech end time” column of the “replacement unit speech data content” column “Saudi Arabia” in the row of the replacement unit ID “1”. The value “0.3” obtained by subtracting the value “2.0” in the “start time” column is stored in the replacement unit audio data length column. Then, the replacement speech synthesizer 22 generates speech data corresponding to the speech uttered “Saudi Arabia” over “0.3” seconds as replacement portion speech data 23 and stores it in the “replacement portion speech data” column. To do. For the second row of the replacement voice data table 24, the same processing as that for the first row is performed.

  In the “shift amount” column of the second embodiment, both the first row and the second row are values calculated from a predetermined formula (replacement unit voice data length− (utterance end time−utterance start time)). Since “0” is entered, the shift process is not performed. FIG. 18 is a diagram illustrating a processing example of the audio data shift processing S6 in the second embodiment, and FIG. 19 is a diagram illustrating a processing example of the audio data replacement processing S7 in the second embodiment. As shown in FIG. 18 and FIG. 19, “Saudi Arabia” replacing “Fukuoka” and “1000” replacing voice data 23 replacing “100 million” are respectively utterance start times of “Fukuoka” and “100 million”. In order not to change the utterance end time, they are generated with the same length without performing shift processing.

  As described above, according to the second embodiment, it is not necessary to process a portion of audio data other than the audio to be replaced, and the audio data replacement processing can be simplified.

(Third embodiment)
In the third embodiment, only the parts different from the first embodiment or the second embodiment will be described. About other points, it is the same as that of 1st Embodiment or 2nd Embodiment.

  FIG. 20 is a diagram illustrating an example of the replacement speech data table 24 and the replacement speech synthesis process S5 in the third embodiment. FIG. 20 is an example in which “Fukuoka” is replaced with “Thailand”, unlike the case where “Fukuoka” is replaced with “Saudi Arabia” in the first and second embodiments. The replacement speech synthesizer 22 is similar to the first embodiment and the second embodiment in that a replacement speech data table 24 is generated and output based on the input specific information utterance time table 21. The replacement speech synthesizer 22 then sets the value “2.0” described in the “utterance start time” column of “Fukuoka” in the “text data before replacement” column of the specific information utterance time table 21 and the “utterance”. The first implementation is that the value “2.3” described in the “End time” column is entered in the “Speech start time” column and the “Speech end time” column of the replacement voice data table 24, respectively. It is the same as that of form and 2nd Embodiment.

  However, the replacement speech synthesizer 22 of the third embodiment describes the “combination ID” column “10” of the specific information utterance time table 21, unlike FIGS. 17 and 18 of the second embodiment. The description item “tie” in the “replaced text data” column in the first row is replaced with the “replacement unit audio data content” in the first row in which the “replacement unit ID” column “1” in the replacement speech data table 24 is described. ”Column. Then, the replacement speech synthesizer 22 of the third embodiment generates replacement portion speech data 23 corresponding to speech in which the entry item “tie” in the “substitution text data” column is spoken at a natural speaking speed. Are stored in the “replacement section voice data length” column, and the value “0.1” as the speech length is entered in the “replacement section voice data length” column.

  FIG. 21 is a diagram illustrating a processing example of the audio data shift processing S6 in the third embodiment. The voice data shift unit 25 generates and outputs shift utterance section voice data 27 based on the input utterance section voice data 15 and the replacement voice data table 24. Since “2.3” is described in the utterance end time in the first row of the replacement voice data table 24 and “0.0” is described in the shift amount column, the voice data shift unit 25 is input. All the voice data after 2.3 seconds of the utterance section voice data 15 is shifted 0.0 seconds in the positive direction.

  FIG. 22 is a diagram illustrating a processing example of the audio data replacement processing S7 in the third embodiment. The voice data replacement unit 26 generates and outputs a replacement history table 28 and conversion voice data 8 based on the input shift utterance section voice data 27 and the replacement voice data table 24. Since “2.3” is described in the “speech end time” column and “0.1” is described in the “replacement unit speech data length” column of the first row of the replacement speech data table, the shift speech segment speech data 27, the voice data from 2.0 seconds to 2.2 seconds is replaced with the voice data corresponding to silence, and the voice data from 2.2 seconds to 2.3 seconds is further replaced with the first row of the replacement voice data table. In the “replacement unit audio data” column of “replacement unit audio data”.

  As described above, according to the third embodiment, processing for audio data other than the audio to be replaced becomes unnecessary, and the audio data replacement processing can be simplified.

  As described above, according to the present disclosure, the possibility of leakage of confidential information to the outside when transmitting utterance voice data to the voice recognition server is reduced, and voice recognition / text conversion is performed with sufficient accuracy. be able to. In addition, the present disclosure can reduce the load of speech recognition / text processing performed by the internal information processing terminal, and can ensure the processing speed.

  The present disclosure reduces the possibility of leakage of confidential information to the outside when transmitting utterance voice data to the voice recognition server when voice data is converted into text, and performs voice recognition / text conversion with sufficient accuracy. Data processing apparatus, data processing system, data processing method, and data that can reduce the load of voice recognition / text processing performed by an internal information processing terminal and can ensure processing speed It is useful as a processing program.

DESCRIPTION OF SYMBOLS 1 Voice processing system 2 The place where voice recognition is performed 3 Voice input processing part 4 Display part 5 Voice recognition speaker 6 Speech voice 7 Data processing device 8 Voice data for conversion 9 Network 10 Voice recognition server 11 Text data 12 Operation part 13 Collection Audio voice data 14 Utterance section detection section 15 Utterance section audio data 16 Specific voice data detection section 17 Utterance time detection section 18 Combination storage section 19 Combination table 20 Specific section voice data 21 Specific information utterance time table 22 Replacement voice synthesis section 23 Replacement section Speech data 24 Replacement speech data table 25 Speech data shift unit 26 Speech data replacement unit 27 Shift utterance interval speech data 28 Replacement history table 29 Replacement history storage unit 30 Communication unit 31 Text data reverse replacement unit 32 Recognition result

Claims (20)

A data processing device that outputs a recognition result of collected sound data,
An audio data replacement unit that replaces the specific unit audio data included in the collected audio data with replacement unit audio data different from the specific unit audio data and outputs the converted audio data;
A communication unit that transmits the voice data for conversion to a voice recognition server and receives text data converted from the voice data for conversion from the voice recognition server;
Of the text data input from the speech recognition server, extract post-replacement text data corresponding to the replacement unit speech data, and replace the post-replacement text data with pre-replacement text data corresponding to the specific unit speech data A text data reverse replacement unit that outputs as a recognition result of the collected sound data;
A data processing apparatus comprising: A combination storage unit for storing a combination of the pre-replacement text data corresponding to the specific unit audio data and the post-replacement text data;
A replacement speech synthesizer that synthesizes the replacement unit speech data from the replaced text data and outputs the synthesized speech data to the speech data replacement unit;
Further comprising
The data processing apparatus according to claim 1. An utterance time detection unit for detecting an utterance start time and an utterance end time of the specific unit audio data;
The speech data shift unit that shifts the speech end time and the speech data after the speech end time according to the speech time of the replacement unit speech data;
Further comprising
The data processing apparatus according to claim 1 or 2. The voice data shift unit sets the utterance start time of the replacement unit voice data as the utterance start time of the specific unit voice data.
The data processing apparatus according to claim 3. The voice data shift unit calculates the utterance end time of the replacement unit voice data from the utterance start time and the utterance time of the replacement unit voice data, and determines the utterance start time of the voice data following the replacement unit voice data. , Shift after the calculated utterance end time,
The data processing apparatus according to claim 4. The voice data shift unit is configured such that when the utterance time of the replacement unit voice data is equal to or shorter than the utterance time of the specific unit voice data, the utterance end time of the replacement unit voice data becomes the utterance end time of the specific unit voice data. , Shifting the utterance start time of the replacement unit voice data,
The data processing apparatus according to claim 3. An operation unit for editing a combination of the pre-replacement text data corresponding to the specific unit audio data and the post-replacement text data;
The data processing apparatus according to claim 1. The audio data shift unit generates the replacement unit audio data having the same length as the specific unit audio data.
The data processing apparatus according to claim 3. The voice data replacement unit further includes a replacement history storage unit that stores a history of replacing the specific unit voice data with the replacement unit voice data.
The data processing apparatus according to claim 1. A data processing system that outputs a recognition result of collected sound data,
An audio data replacement unit that replaces the specific unit audio data included in the collected audio data with replacement unit audio data different from the specific unit audio data and outputs the converted audio data;
A communication unit that transmits the voice data for conversion to a voice recognition server and receives text data converted from the voice data for conversion from the voice recognition server;
Of the text data input from the speech recognition server, extract post-replacement text data corresponding to the replacement unit speech data, and replace the post-replacement text data with pre-replacement text data corresponding to the specific unit speech data A text data reverse replacement unit that outputs as a recognition result of the collected sound data;
A data processing system comprising: A data processing method for outputting a recognition result of collected sound data,
A voice data replacement step of replacing the specific part voice data included in the collected voice data with replacement voice data different from the specific part voice data and outputting the voice data for conversion;
A communication step of transmitting the voice data for conversion to a voice recognition server and receiving text data converted from the voice data for conversion from the voice recognition server;
Of the text data input from the speech recognition server, extract post-replacement text data corresponding to the replacement unit speech data, and replace the post-replacement text data with pre-replacement text data corresponding to the specific unit speech data A text data reverse replacement step for outputting the collected sound data as a recognition result;
A data processing method comprising: A combination storage step of storing a combination of the pre-replacement text data corresponding to the specific speech data and the post-replacement text data;
A replacement speech synthesis step of synthesizing the replacement unit speech data from the post-substitution text data and outputting the synthesized speech data to the speech data replacement unit;
Further comprising
The data processing method according to claim 11. An utterance time detection step of detecting an utterance start time and an utterance end time of the specific part voice data;
A speech data shift step for shifting the speech end time and the speech data after the speech end time according to the speech time of the replacement unit speech data;
Further comprising
The data processing method according to claim 11 or 12. In the voice data shift step, the utterance start time of the replacement unit voice data is set as the utterance start time of the specific unit voice data.
The data processing method according to claim 13. The voice data shift step calculates the utterance end time of the replacement unit voice data from the utterance start time and the utterance time of the replacement unit voice data, and sets the utterance start time of the voice data following the replacement unit voice data. , Shift after the calculated utterance end time,
The data processing method according to claim 14. In the voice data shift step, when the utterance time of the replacement part voice data is equal to or shorter than the utterance time of the specific part voice data, the utterance end time of the replacement part voice data becomes the utterance end time of the specific part voice data. , Shifting the utterance start time of the replacement unit voice data,
The data processing method according to claim 13. An operation step of editing a combination of the pre-replacement text data corresponding to the specific part audio data and the post-replacement text data;
The data processing method according to claim 11. The voice data shift step generates the replacement unit voice data having the same length as the specific unit voice data.
The data processing method according to claim 13. The voice data replacement step further includes a replacement history storage step of storing a history of replacing the specific unit voice data with the replacement unit voice data.
The data processing method according to claim 11. A data processing program executed in a data processing device that outputs a recognition result of collected sound data,
For the computer of the data processing device,
A process of replacing the specific unit audio data included in the collected audio data with replacement unit audio data different from the specific unit audio data and outputting the converted audio data as conversion audio data;
Processing for transmitting the voice data for conversion to a voice recognition server and receiving text data converted from the voice data for conversion from the voice recognition server;
Of the text data input from the speech recognition server, the post-replacement text data corresponding to the replacement unit speech data is extracted, and the post-replacement text data is replaced with the pre-replacement text data corresponding to the specific unit speech data. Processing to output the collected sound data as a recognition result;
Data processing program that executes

Images