JP5713782B2 - Information processing apparatus, information processing method, and program - Google Patents

Description

  The present invention relates to a technique for specifying a voice related to a specific part in a meeting minutes, for example.

  2. Description of the Related Art Conventionally, a technique for identifying a location related to an input keyword by keyword matching from a meeting minutes transcribed by voice recognition is known (see, for example, Patent Document 1).

JP 2002-99530 A

  When confirming the details of the contents written in the minutes, the voice during the meeting is recorded and the voice is listened to. And since the user needs a lot of time to listen to all the recorded sounds, it is necessary to listen only to the sounds related to a specific part in the minutes. However, in the keyword matching as disclosed in Patent Document 1, it is difficult to specify related speech when the input keyword exists in a wide range in the minutes.

  Therefore, an object of the present invention is to specify, for example, a voice related to a specific part in a meeting minutes with high accuracy.

An information processing apparatus according to the present invention includes a conversion means for converting each of a plurality of speech sections into a plurality of first text data, with each speech unit of speech data including a plurality of speeches by a plurality of speakers as speech sections, and each speech to section showing a first specifying means for specifying a speech start time及beauty onset words person, a second text data representing a summary of the talk unit, a speaker corresponding to the second text data a first receiving means for receiving the information, the input of a second specifying means for specifying during the input mode of the second text data, and the previous SL first text data Taso respectively, said first perform a second text data, the text matching, matching means for identifying a corresponding portion, the corresponding portion, the input time, the talk start time, information indicating the speaker is identified by said first specifying means Based on the speaker, and having a third specifying means for specifying the voice interval corresponding to the second text data.

  ADVANTAGE OF THE INVENTION According to this invention, it becomes possible to specify the audio | voice relevant to the specific location in the minutes of a meeting, for example with high precision.

1 is a diagram showing an overview of an information processing system according to an embodiment of the present invention. It is a figure which shows the functional structure of the information processing system which concerns on embodiment of this invention. It is a figure which shows the hardware constitutions of the information processing system which concerns on embodiment of this invention. It is a flowchart which shows the process which specifies the audio | voice area (related audio | voice area) relevant to the summary sentence of the minutes. It is a figure which shows the utterance content of the audio | voice recorded during the meeting. It is a figure which shows the result of having specified the speech audio | voice unit and the speech start time from the recorded audio | voice in step S401. It is a figure which shows the result by which the speaker of the speech voice unit was specified in step S402. It is a figure which shows the number of the matching parts in each speech audio | voice unit. It is a flowchart which shows the detail of step S403. It is a figure which shows the example of the minutes produced by the minutes clerk, the summary sentence specified from the minutes, the speaker, and the summary sentence input time. It is a figure which shows the example which the stroke time holding | maintenance part hold | maintained the stroke time with respect to the summary sentence "It is worth giving at a meeting." Shown to 701 of FIG.7 (b). It is a figure which shows the result of having carried out the morphological analysis of the summary sentence "It is worth giving at a meeting." Shown to 701 of FIG. It is a flowchart which shows the detail of the process of step S405. It is a figure for demonstrating the example of reproduction | regeneration of the speech audio | voice unit. It is a flowchart which shows the process in 2nd Embodiment. It is a flowchart which shows the process in 3rd Embodiment. It is a flowchart which shows the process in 4th Embodiment.

  DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments to which the invention is applied will be described in detail with reference to the accompanying drawings.

  First, a first embodiment of the present invention will be described. FIG. 1 is a diagram showing an overview of an information processing system according to the first embodiment of the present invention. The information processing system according to the present embodiment includes a voice conversion function and a speaker specifying function. In addition, the information processing system according to the present embodiment is effective due to the cooperation of the processing units. However, the present invention is not limited to this, and an information processing device integrally including all processing units is also applicable to the present invention. It is.

  The information processing system shown in FIG. 1 mainly includes a microphone 101 and a PC 102. When the conference is performed by the speakers 103 to 107, the microphone 101 records the voices spoken by the speakers 103 to 107. The minutes clerk 108 uses the PC 102 to input a summary of the content of the statement and create a meeting minutes. Here, a set of summary sentences is used as the minutes of the meeting. Further, for convenience of explanation, the names of the speakers 103 to 107 are hereinafter referred to as “Sato”, “Tanaka”, “Suzuki”, “Ito”, and “Okawa”.

  FIG. 2 is a diagram illustrating a functional configuration of the information processing system according to the present embodiment. In FIG. 2, the voice conversion unit 201 converts the utterance content recorded by the microphone 101 into text. The speaker specifying unit 202 specifies a speaker of the content of the speech recorded by the microphone 101. The input unit 203 has a configuration corresponding to the PC 102, and a summary sentence and a speaker of the conference speech content are input by the minutes clerk 108. The matching part specifying unit 204 performs text matching between the utterance content and the summary sentence, and specifies a part that matches each other (hereinafter referred to as a matching part). The speech section specifying unit 205 specifies a speech section related to the summary sentence using the speech time, the summary sentence input time, the matching portion, and the information of the speaker. The input unit 203 includes a text input unit 206 that inputs a summary sentence as text, a stroke time holding unit 207 that holds a stroke time of the summary sentence, and an input time specification that specifies a time when the summary sentence is input using the stroke time. The unit 208 is provided. In the following description, a voice section related to the summary sentence of the minutes may be referred to as a related voice section.

  FIG. 3 is a diagram illustrating a hardware configuration of the information processing system according to the present embodiment. The CPU 301 operates according to a program so as to realize each operation procedure of the present embodiment. The RAM 302 provides a storage area necessary for the operation of the program. The ROM 303 holds a program, a database, and the like that realize the operation procedure of the program. The voice input device 304 corresponds to the microphone 101 and records voices spoken by the speakers 103 to 107. The text input device 306 corresponds to the PC 102 and inputs a summary sentence of the content of the utterance according to the operation of the minutes clerk 108. The audio playback device 305 outputs audio corresponding to the related audio segment specified by the audio segment specifying unit 205. The summary sentence input by the text input device 306 is given time information when the summary sentence is input. Each processing unit exchanges data via the bus 307.

  FIG. 4 is a flowchart showing a process of specifying a voice section (related voice section) related to the summary sentence of the minutes. Hereinafter, the processing of the information processing system according to the present embodiment will be described with reference to FIG.

  In step S401, the voice segment identification unit 205 detects the voice that actually made a speech (hereinafter referred to as a speech voice unit) from the recorded voice, and the time when the speech voice unit was started (hereinafter referred to as speech start). (Referred to as time). The recorded voice is, for example, WAVE data sampled at 22.05 KHz, and is stored in the external storage device by the PCM method. In the present embodiment, a technique for detecting a voice section is used for detecting a speech voice unit. Here, the speech corresponding to the speech segment detected by the speech segment detection technique is set as the speech unit. In addition, since the technique of speech area detection is well-known, detailed description is abbreviate | omitted.

  In addition, the recorded time is written in the header portion of the WAVE data, and the speech section specifying unit 205 specifies the speech start time from this information. The voice conversion unit 201 converts a speech voice unit into text. A speech recognition technique is used to convert the speech unit into text. In the present embodiment, speech recognition is performed by previously storing speech data of vocabulary related to various meetings in the RAM 302 as a model. In addition, since the technique of voice recognition is well-known, detailed description is abbreviate | omitted. The text generated by the voice conversion unit 201 is an application example of the first text data.

  FIG. 5A is a diagram showing the content of speech recorded during a conference. FIG. 5B is a diagram illustrating a result of specifying the speech voice unit and the speech start time from the recorded voice in step S401. That is, in the column of “speech speech unit” in FIG. 5B, specific results of speech speech units are listed. Further, in the column of “speech start time” in FIG. 5B, the speech start time for each speech sound unit is listed. In the column of “speech content conversion result” in FIG. 5B, the results of converting each speech unit into text are listed. Current speech recognition technology is not accurate enough to completely convert speech to text. Therefore, misrecognition occurs as shown in the column “conversion result of speech” in FIG. 5B. Note that a voice that has no text candidate to be converted is blank. Moreover, in the following description, the result of converting a speech unit into text may be referred to as a speech content conversion result.

  In step S402, the speaker identifying unit 202 identifies the speaker in units of speech voice detected in step S401. In this embodiment, a speaker recognition technique is used to specify a speaker. In the speaker recognition technique, the voice characteristics of the speakers 103 to 107 are stored in the RAM 302 in advance as a model, and the speaker is recognized by collating the voice feature amount obtained from the speech voice unit with the model. . In addition, since the technique of speaker recognition is well-known, detailed description is abbreviate | omitted. FIG. 5C is a diagram illustrating a result of identifying a speaker in units of speech in step S402. That is, the “speaker” column in FIG. 5C lists the speakers in units of speech voices identified in step S402. Note that the speech start time specifying processing in step S401 and the speaker specifying processing in step S402 are processing examples of the first specifying means.

  In step S403, the minutes clerk 108 inputs a summary sentence and a speaker from the input unit 203. Here, step S403 will be described in detail with reference to FIG. FIG. 6 is a flowchart showing details of step S403.

  In step S <b> 601, the text input unit 206 of the input unit 203 inputs a summary sentence of the utterance content and a speaker as text in accordance with the operation of the minutes clerk 108. FIG. 7A shows an example of the minutes created by the minutes clerk 108 inputting text. In the present embodiment, a sentence following the symbol “•” is specified as a summary sentence, and a character string surrounded by symbols “(” and “)” at the end of the summary sentence is used as a speaker of the summary sentence. Identified. FIG. 7B shows a summary sentence and a speaker identified from the created minutes. Note that the text input unit 206 has a configuration as an application example of the first input unit, and the text input by the text input unit 206 is second text data.

  In step S602, the stroke time holding unit 207 of the input unit 203 holds the stroke time of the summary sentence input in step S601. The stroke time holding unit 207 in the present embodiment records one character constituting the summary sentence as a stroke unit, and records the time (stroke time) when the input of one character in each stroke unit is started. FIG. 8 is a diagram showing an example in which the stroke time holding unit 207 holds the stroke time for the summary sentence “It is worth presenting at the meeting” shown at 701 in FIG. 7B.

  In step S603, the input time specifying unit 208 of the input unit 203 specifies the time when the input of the summary sentence is started from the result of step S602 (hereinafter referred to as the summary sentence input time). The stroke time in the first stroke unit of each summary sentence is the summary sentence input time. For the summary sentence “It is worth presenting at the meeting.”, “8:04:50”, which is the stroke time of the “meeting” in units of strokes, is the summary sentence input time. In the column of “summary sentence input time” in FIG. 7C, the summary sentence input times specified for each summary sentence input in step S601 are listed. Therefore, in step S403, for example, when the minutes shown in FIG. 7A are input, the summary sentence input time, the summary sentence, and the speaker are specified as shown in FIG. 7C. The input time specifying unit 208 is a configuration serving as an application example of the second specifying unit.

  In step S404, the matching part specifying unit 204 performs text matching between the text converted in step S401 and the summary sentence input in step S403, and specifies a matching part. Here, the text matching will be specifically described. First, the matching part specifying unit 204 performs morphological analysis on the summary sentence in FIG. FIG. 9 is a diagram showing a result of a morphological analysis of the summary sentence “It is worth presenting at a meeting” shown at 701 in FIG. 7B. The summary sentence 701 is divided into words 1 to 8 and the part of speech of each word is specified. Then, the matching part specifying unit 204 searches the words “conference”, “announcement”, and “value” in which the part of speech is specified as a noun from the result of conversion of the content of the speech in FIG. 5C. The matching location specifying unit 204 sets the location of the searched word as a matching location. Note that a concept dictionary or the like may be used as an alternative means of text matching, and a word part having a close meaning may be used as a matching part. FIG. 5D is a diagram showing the number of matching points in each speech unit.

  In step S405, the speech section specifying unit 205 uses the speech start time, the summary sentence input time, the matching part, and the information of the speaker to determine a speech section (related speech section) related to the summary sentence input in step S403. Identify. Here, step S405 will be described in detail with reference to FIG. FIG. 10 is a flowchart showing details of the process in step S405. Hereinafter, an example in which a speech section (related speech section) related to the summary sentence 701 in FIG. 7B is specified will be described with reference to FIG. Note that the speech segment identification unit 205 is a configuration that is an application example of the third identification unit.

  In step S1001, the speech section specifying unit 205 specifies a time information target section from the results of steps S401 and S403. Here, from the summary sentence input time (8:04:50) of the summary sentence 701 in FIG. 7B, speech speech units within a predetermined time (here, 2 minutes) are defined as the time information target section. To do. That is, in FIG. 5C, speech speech units 511 to 515 between 8:04:50 and 8:02:50 among speech speech units 501 to 517 are time information target sections. In step S1002, the speech section specifying unit 205 specifies the speaker information target section from the results of steps S402 and S403. Here, the speech unit that is spoken by the speaker (Sato) input in the summary sentence 701 is set as the speaker information target section. That is, in FIG. 5C, among the speech units 501 to 517, speech units 501, 505, 513, and 515 uttered by the speaker (Sato) are the speaker information target sections.

  In step S1003, the speech section specifying unit 205 specifies the matching part information target section from the result of step S404. Here, a speech unit whose number of matching points in each speech unit is equal to or more than a first threshold (here, 2) is set as a matching part information target section. That is, in FIG. 5D, speech speech units 503, 505, 507, 513, 515, and 516 whose matching locations are equal to or greater than the first threshold among speech speech units 501 to 517 are identified as matching location information target sections.

  In step S1004, the speech section specifying unit 205 specifies a related speech section from the results of steps S1001 to S1003. Here, the speech information unit in which the time information target section, the speaker information target section, and the matching location information target section overlap (become AND) is set as the related speech section. That is, in step S405, speech voice units 513 and 515 are specified as related voice sections.

  Therefore, by using the time information and the speaker information in addition to the keyword matching by the processing of step S401 to step S405, it is possible to specify the speech section related to the summary sentence in the minutes with high accuracy. As a specific application of the information processing system, there is a case where the details of the summary sentence written in the minutes are desired to be played back after the meeting is finished. For example, in FIG. 11, when the minutes are displayed on the screen 1102 of the PC 1101 and the portion 1104 corresponding to the summary sentence 701 in FIG. 7 is clicked with the mouse 1103, the speech units 513 and 515 are reproduced. In the information processing system described above, a conference has been described as an example. However, the present invention can be applied to any form such as a lecture or a class as long as the content of a speaker can be recorded.

  Next, a second embodiment of the present invention will be described. In the second embodiment, the processing shown in FIG. 12 is executed after step S1004 of FIG. 10 in order to determine whether or not the speech unit adjacent to the related speech section is included in the related speech section. Hereinafter, the process illustrated in FIG. 12 will be described. Note that the configuration of the information processing system according to the second embodiment is the same as the configuration of the information processing system according to the first embodiment, and therefore will be described using the same reference numerals as those in the first embodiment.

  In step S1201, the speech segment specifying unit 205 sets the number of matching points obtained in step S404 in the speech unit adjacent to the identified related speech segment (hereinafter referred to as the adjacent speech unit). It is determined whether or not it is equal to or greater than a threshold value (here, 2). If the number of matching points is equal to or greater than the second threshold, the process proceeds to step S1202. On the other hand, if the number of matching points is less than the second threshold, the process ends. In step S1202, the speech segment specifying unit 205 includes the adjacent speech unit in the specified related speech segment. In the present embodiment, if the number of matching points in the adjacent speech unit 516 is two or more, the adjacent speech unit 516 is included in the related speech section. Therefore, even if it is not specified as the related speech section in step S1004, if the number of matching points that are likely to be highly relevant is equal to or greater than the second threshold, the adjacent speech unit is included in the related speech section.

  Next, a third embodiment of the present invention will be described. In the third embodiment, when there are a plurality of related speech sections of the summary sentence 701 in FIG. 7B specified in step S405, the process shown in FIG. 13 is executed after step S1004. Hereinafter, the process illustrated in FIG. 13 will be described. Note that the configuration of the information processing system according to the third embodiment is the same as the configuration of the information processing system according to the first embodiment, and therefore will be described using the same reference numerals as those in the first embodiment.

  In step S1301, the speech segment identification unit 205 determines whether there are a plurality of identified related speech segments. If there are a plurality of related speech sections, the process proceeds to step S1302. On the other hand, when there are not a plurality of related speech sections, the process ends. For the summary sentence 701, the speech voice units 513 and 515 are specified as the related voice section, and the process proceeds to step S1302.

  In step S <b> 1302, the speech segment specifying unit 205 lowers the second threshold set for the adjacent speech speech units located between the plurality of specified related speech segments. Here, it is assumed that the second threshold value of the adjacent speech unit 514 is lowered from 2 to 1. In step S1303, the speech section specifying unit 205 determines whether or not the number of matching points obtained in step S404 is equal to or greater than the second threshold set in step S1302 in adjacent speech units. If the number of matching points is equal to or greater than the second threshold, the process proceeds to step S1304. On the other hand, if the number of matching points is less than the second threshold, the process ends. Since the adjacent speech unit 514 is between a plurality of specified related speech sections, the second threshold is set to 1. Accordingly, for the adjacent speech unit 514, the number of matching points is 1, so the process moves to step S1304. Moreover, since the number of matching parts of the adjacent speech unit 512 is less than the second threshold value, the process ends. On the other hand, since the number of matching points in the adjacent speech unit 516 is equal to or greater than the second threshold, the process proceeds to step S1304.

  In step S1304, the speech segment specifying unit 205 includes the adjacent speech unit in the related speech segment specified in step S1004. Therefore, even if it is not specified as the related speech section in step S1004, the adjacent speech sound units between the related speech sections are included in the related speech section with a higher probability.

  Next, a fourth embodiment of the present invention will be described. In the fourth embodiment, when there are a plurality of speakers in the summary sentence 701 in FIG. 7B input in step S403, the processing shown in FIG. 14 is executed after step S1004 in FIG. Hereinafter, the process illustrated in FIG. 14 will be described. Note that the configuration of the information processing system according to the fourth embodiment is the same as the configuration of the information processing system according to the first embodiment, and therefore will be described using the same reference numerals as those in the first embodiment.

  In step S1401, the speech section identification unit 205 determines whether there are a plurality of speakers in the summary sentence input in step S403. If there are a plurality of speakers in the summary sentence, the process proceeds to step S1402. On the other hand, if there are not a plurality of speakers in the summary sentence, the process ends. In step S1402, the speech segment specifying unit 205 determines whether or not the speaker in the adjacent speech unit of the specified related speech segment is the speaker of the summary sentence input in step S403. If the speaker is an input summary sentence, the process proceeds to step S1403. On the other hand, if the speaker is not a speaker of the input summary sentence, the process ends. In step S1403, the speech segment specifying unit 205 includes the adjacent speech unit in the related speech segment specified in step S1004.

  For example, in the minutes shown in Fig. 7 (a), "・ It is worth making a presentation at a meeting (Sato)" is written as "・ It is worth making a presentation at a meeting (Sato, Suzuki)". Suppose. The speakers of the summary sentence 701 are Sato and Suzuki. A character string surrounded by “(” and “)” symbols and separated by “,” symbols is specified as a speaker of the summary sentence. The speaker information target section specified in step S1002 is speech voice units 501, 503, 505, 506, 508, 510, 513, 515, and 516. The result of step S1004 (speech voice units 513 and 515) is does not change. However, since the speaker of the adjacent speech unit 516 is Suzuki, the adjacent speech unit 516 is included in the related speech section. Therefore, even if it is not specified as the related speech section in step S1004, the speech section related to the speaker entered by the minuteskeeper can be specified with high accuracy. Therefore, even if it is not specified as the related voice section in step S1004, the adjacent speech voice unit of the speaker entered by the minutes clerk who may be highly relevant is included in the related voice section.

  The present invention can also be realized by executing the following processing. That is, software (program) that realizes the functions of the above-described embodiments is supplied to a system or apparatus via a network or various storage media, and a computer (or CPU, MPU, or the like) of the system or apparatus reads the program. It is a process to be executed.

  201: Voice conversion unit, 202: Speaker specifying unit, 203: Input unit, 204: Matching part specifying unit, 205: Voice segment specifying unit, 206: Text input unit, 207: Stroke time holding unit, 208: Input time specifying Part

Claims (13)

Conversion means for converting each of the plurality of speech sections into a plurality of first text data, with a speech unit of speech data including a plurality of speeches by a plurality of speakers as a speech section ;
For each speech segment, and the first specifying means for specifying a speech start time及beauty onset words person,
First accepting means for accepting input of second text data representing a summary sentence of the comment unit, and information indicating a speaker corresponding to the second text data ;
Second specifying means for specifying during the input mode of the second text data,
Before SL first text data Taso Rezoreto performed with the second text data, the text matching, matching means for identifying a corresponding portion,
The speech section corresponding to the second text data is identified based on the corresponding location, the input time, the speech start time, information indicating the speaker, and the speaker identified by the first identifying means. And an information processing apparatus. Second accepting means for accepting designation of the summary sentence of the second text data;
Output means for outputting the voice data of the voice section specified by the third specifying means for the second text data when the designation is received;
The information processing apparatus according to claim 1, further comprising: The third specific means, as the object of the speech segment in which the speech start time within the from the input time of the predetermined time is determined, the speech section corresponding to the second text data from the target The information processing apparatus according to claim 1, wherein the information processing apparatus is specified. The third specifying means specifies the speech section corresponding to the second text data from the target, with the speech section in which the speaker indicated by the information indicating the speaker is specified as a target. The information processing apparatus according to claim 1, wherein the information processing apparatus is an information processing apparatus. The third specifying means targets the speech section corresponding to the first text data in which the number of the corresponding locations specified by the matching means is equal to or more than a first threshold, from among the objects . 5. The information processing apparatus according to claim 1, wherein the speech section corresponding to two text data is specified . The third specifying means specifies the speaker start time within a predetermined time from the input time, specifies the speaker who matches the speaker information, and the number of corresponding locations is a first threshold value. The information processing apparatus according to claim 1, wherein the speech section corresponding to the first text data is specified as the speech section corresponding to the second text data . The third specifying means specifies the voice section, and the corresponding portion between the first text data and the second text data corresponding to an adjacent voice section adjacent to the specified voice section is the first. 7. The information processing apparatus according to claim 1, wherein when the threshold value is equal to or greater than 2, the adjacent speech section is included in the speech section corresponding to the second text data . The third specifying unit is adjacent to at least one voice section of the plurality of voice sections and has the plurality of voice sections when a plurality of voice sections are specified for the second text data. The speech section located between the first text data and the second text data corresponding to the speech section to be processed is smaller than the second threshold value. 8. The information processing apparatus according to claim 7, wherein the processing target speech section is included in the speech section corresponding to the second text data when the third threshold value or more. The first accepting means accepts input of information indicated by the speaker, in which a plurality of speakers are indicated,
  The third specifying means, when the voice section is specified and the speaker specified for the adjacent voice section adjacent to the specified voice section is included in the information indicating the speaker, The information processing apparatus according to claim 1, wherein an adjacent speech section is included in the speech section corresponding to the second text data. The said 2nd specific means specifies the said input time based on the information of the stroke time concerning the input of said 1st text data, The any one of Claim 1 thru | or 9 characterized by the above-mentioned. Information processing device. An information processing method executed by an information processing apparatus,
A conversion step of converting each of the plurality of speech sections into a plurality of first text data, with a speech unit of the speech data including a plurality of comments by a plurality of speakers as a speech section;
A first identification step for identifying a speech start time and a speaker for each voice section;
A first accepting step for accepting input of second text data representing a summary sentence of the comment unit, and information indicating a speaker corresponding to the second text data;
A second specifying step of specifying an input time of the second text data;
A matching step of performing text matching between each of the first text data and the second text data to identify a corresponding portion;
The speech section corresponding to the second text data is identified based on the corresponding location, the input time, the speech start time, information indicating the speaker, and the speaker identified by the first identifying means. A third specific step to
An information processing method comprising: Computer
Conversion means for converting each of the plurality of speech sections into a plurality of first text data, with a speech unit of speech data including a plurality of speeches by a plurality of speakers as a speech section;
A first specifying means for specifying a speech start time and a speaker for each voice section;
First accepting means for accepting input of second text data representing a summary sentence of the comment unit, and information indicating a speaker corresponding to the second text data;
Second specifying means for specifying an input time of the second text data;
Matching means for performing text matching between each of the first text data and the second text data, and identifying a corresponding portion;
The speech section corresponding to the second text data is identified based on the corresponding location, the input time, the speech start time, information indicating the speaker, and the speaker identified by the first identifying means. Third identifying means to
Program to function as. A computer-readable recording medium storing the program according to claim 12.

Images