JP2004279897A - Method, device, and program for voice communication record generation - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide the minutes of a remote communication conference whose main point is made clear under conditions that a person who requires the minutes desires when the minutes are provided. <P>SOLUTION: A method for generating voice communication record includes the stages of: recognizing speech information from a speech storage means of storing speech information from at least two or more points together with time information when the speech information is inputted and converting it into text information; deciding information on the time when the text information includes at lest one topic keyword from a topic keyword storage means storing top keywords or its synonym and recording the time information and topic keyword in a recording means while making them correspond to each other; deciding time information when the topic keyword or its synonym is recorded in the recording medium among pieces of inputted retrieval request information; and outputting the decided time information or speech information corresponding to a specified time section including the time information from the speech storage means. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a voice communication record generation method, an apparatus, and a program. In particular, the present invention relates to a method for creating a record that allows a user to easily retrieve the contents of acquired voice information and the like in a telecommunication conference using the Internet.
[0002]
[Prior art]
2. Description of the Related Art Conventionally, in a conference system using multimedia information such as audio or images, there has been a device that can easily edit important items during a conference in a short time (for example, see Patent Document 1).
[0003]
This apparatus analyzes the temporal relationship of multimedia information obtained by a conference, identifies a speaker, and extracts a keyword from the multimedia information. In particular, for voice information, keywords are extracted by voice recognition processing. The speaker is identified by the device identification number to which the voice information is input and the feature extraction process. Further, the apparatus accumulates a search file indicating the importance of the conference in a time series by weighting the keyword and the speaker. These processes are performed in real time during the meeting. This makes it possible to narrow down scenes containing important items on the basis of keywords, reference materials, material referrers or speakers, time periods of meetings, and the like that appear during the meeting, and the time required for creating meeting minutes Can be greatly reduced.
[0004]
In addition, as a mechanism for supporting minutes creation, mixed data in which voice information of multiple speakers is mixed is separated into voice information for each speaker, and the voice minutes are created by specifying the speaker. There was also a method (for example, see Patent Document 2).
[0005]
Further, a consultation conference system that records not only audio and video but also graphic information and sends out only essential points in response to requests has been realized (for example, see Patent Document 3).
[0006]
[Patent Document 1]
Japanese Patent No. 3185505 [Patent Document 2]
JP-A-2003-5790 (Japanese Patent Application No. 2001-191289)
[Patent Document 3]
JP-A-5-298340 (Japanese Patent Application No. 4-96883)
[0007]
[Problems to be solved by the invention]
According to Patent Literature 1, it is assumed that the creation of a conference record is performed by a search creator inputting search conditions. In particular, if the desired data cannot be reached, a request is made to the transcript creator to input further search conditions. Therefore, it is possible to create only one detailed minutes by the minutes creator. However, the desired points of the minutes are often different for each conference participant. That is, according to Patent Literature 1, different conference minutes cannot be provided to each conference participant.
[0008]
In addition, according to Patent Literature 2, there is a barrier of accuracy in speaker identification, and according to Patent Literature 3, there is also a disadvantage that there is no index function and lacks convenience.
[0009]
That is, according to the conventional technique, the conference minutes are prepared only for the convenience of the conference creator, and the conference minutes in which the main points are summarized according to the desired conditions of the person who needs the conference minutes are not provided.
[0010]
In view of the above, an object of the present invention is to provide a conference record in which the main points are summarized according to desired conditions of a person who needs the conference record.
[0011]
[Means for Solving the Problems]
The present invention relates to a telecommunications conference, a method for generating a voice communication record using two or more terminals used by each participant, and a voice communication record generation device for recording voice information in the conference and generating a meeting record. , Devices and programs.
[0012]
According to the voice communication record generation method according to the present invention,
Voice information from at least two or more points, and a voice storage means for storing the voice information together with the input time information, voice-recognizing the voice information and converting it into text information;
Determining time information in which the text information includes at least one of the topic keywords or a synonym thereof from the topic keyword storage unit that accumulates the topic keywords, and records the time information in the recording unit in association with the topic keywords;
A step of determining time information at which the topic keyword or a synonym thereof is recorded in the recording unit of the input search request information;
Outputting the determined time information or voice information corresponding to a predetermined time section including the time information from the voice storage means.
[0013]
Further, according to another embodiment of the voice communication record generation method of the present invention, the method may include a step of detecting a speech section and a silent section of voice information.
[0014]
Further, according to another embodiment of the voice communication record generation method of the present invention, a synonym is further searched for by using a synonym storage unit that previously stores synonyms for the topic keyword, and the progress position is marked on the text information together with the topic keyword. It is also preferable to do so.
[0015]
Further, according to another embodiment of the voice communication record generation method of the present invention, it is preferable that the conference information is added to the voice information, and the voice information is linked to other information.
[0016]
According to the voice communication record generation device of the present invention,
Means for converting voice information into text information by voice recognition from voice storage means for storing voice information from at least two or more points and time information at which the voice information is input;
Means for determining time information including at least one of the topic keywords or synonyms thereof from the topic keyword storage means for accumulating topic keywords in the text information, and recording the time information and the topic keywords in the recording means in association with each other;
Means for determining the time information at which the topic keyword or a synonym thereof is recorded in the recording means of the input search request information;
Means for outputting the determined time information or voice information corresponding to a predetermined time section including the time information from the voice storage means.
[0017]
According to another embodiment of the voice communication record generation device of the present invention, the voice communication record generation device may include a unit for detecting a speech section and a silent section of voice information.
[0018]
Further, according to another embodiment of the voice communication record generation device of the present invention, a synonym is further searched for by using a synonym storage means in which a synonym for the topic keyword is stored in advance, and the progress position is marked on the text information together with the topic keyword. It is also preferable to do so.
[0019]
Further, according to another embodiment of the audio communication record generation device of the present invention, the conference information may be added to the audio information, and the audio information may be linked to other information.
[0020]
According to the voice communication record generation program of the present invention,
Voice information from at least two or more points, and a voice storage means for storing the voice information together with the input time information, voice-recognizing the voice information and converting it into text information;
Determining time information in which the text information includes at least one of the topic keywords or a synonym thereof from the topic keyword storage means for accumulating the topic keywords, and recording the time information in the recording means in association with the topic keywords;
A step of determining time information at which the topic keyword or a synonym thereof is recorded in the recording unit of the input search request information;
The computer is executed as a process of outputting the determined time information or voice information corresponding to a predetermined time section including the time information from the voice storage unit.
[0021]
According to another embodiment of the voice communication record generation program of the present invention, the computer may be executed so as to have a process of detecting a speech section and a silent section of voice information.
[0022]
Further, according to another embodiment of the voice communication record generation program of the present invention, a synonym is further searched for by using a synonym storage unit that previously stores synonyms for the topic keyword, and the progress position is marked on the text information together with the topic keyword. It is also preferable to execute the computer in such a manner as to execute the above.
[0023]
Furthermore, according to another embodiment of the voice communication record generation program of the present invention, the conference specific information is added to the voice information, and the computer is executed such that the voice information is linked to other information. Is also preferred.
[0024]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[0025]
FIG. 1 is a system configuration diagram according to the present invention.
[0026]
According to FIG. 1, a voice communication recording server 1 according to the present invention is connected to participants' terminals 2, 3 and 4 via the Internet 5. The conference participants can hold a telecommunication conference via the Internet 5 using the respective terminals 2, 3 and 4. Here, participants with participant IDxxx, yyy, and zzz are participating in the telecommunication conference.
[0027]
Each of the terminals 2, 3 and 4 includes a microphone 21, 31 and 41 for acquiring audio information, and video cameras 22, 32 and 42 for acquiring video information.
[0028]
The exchange of voice information and the like between terminals in the conference may be delivered via the voice communication recording server 1 or may be broadcast to the voice communication recording server 1 and the terminals of the conference participants. It may be distributed.
[0029]
The audio information or the like may be one in which the audio information or the like is stored in a file format by each terminal from the start to the end of the conference. In this case, on the other hand, there is a method of transmitting a file of audio information recorded by each of the terminals 2, 3 and 4 to the audio communication recording server 1 at a time after the conference. On the other hand, there is a method of sequentially transmitting voice information or the like in a file format for each utterance to the voice communication recording server 1 during a conference. By classifying voice information and the like for each utterance, it is possible to respond to detailed search conditions. At this time, it is also preferable that each terminal detects a silent section and transmits only the voice information of the voiced section to the voice communication recording server 1.
[0030]
On the other hand, the audio information or the like may be transmitted from each terminal to the audio communication recording server 1 in a streaming format from the start to the end of the conference. In this case, in the voice communication recording server 1, voice information and the like are classified according to each speaker and / or each utterance. At this time, it is also preferable that the audio communication recording server 1 detects a silent section and extracts only audio information of a sound section.
[0031]
It should be noted that the present invention intends to edit and search the contents of the meeting after the meeting. Therefore, in the embodiment, a communication conference by a plurality of terminals is assumed, but it is also realistic to aggregate voice information and the like in one device and perform processing by a computer.
[0032]
FIG. 2 is a functional configuration diagram of the voice communication recording server 1 according to the present invention.
[0033]
According to FIG. 2, the voice communication recording server 1 includes a voice information database 10 (voice storage means), a video information database 11, a time stamp 12, a conference ID determination unit 13, a voice recognition processing unit 14, It has a minutes information database 15 (recording means), a gist search unit 16, a minutes generation unit 17, a synonym database 18, and a communication interface 19.
[0034]
The Internet 5 is connected via a communication interface 19.
[0035]
The conference ID determination unit 13 receives the conference start registration message including the “conference name”, “participant name”, and the like of the conference, and determines the corresponding “conference ID” and “participant ID” (conference specific information). , And deliver to the participants' terminals. These IDs can be attached to audio information and video information received from the terminal.
[0036]
The time stamp unit 12 attaches a time to the audio information and the video information received by the communication interface 19.
[0037]
The received voice information is stored in the voice information database 10 for each participant ID. The video information is stored in the video information database 11 for each participant ID. Here, the audio information and the video information are linked by time using a time stamp. Note that the video information is not limited to information obtained from a video camera provided in a terminal used by a participant, but is material data having a visual effect such as document material data or projection material data used in a meeting record. You may.
[0038]
Further, the voice information database 10 detects an utterance section and a silent section in the voice information, and accumulates voice information of only the utterance section. As a method for realizing this, there is, for example, Japanese Patent No. 2590193.
[0039]
Next, the voice information stored in the voice information database 10 is converted into text information in the voice recognition processing unit 14. At this time, for the text information, a “feature word” meaning the progress position of the topic of the meeting is searched. Further, a “topic keyword” corresponding to the “feature word” is searched. Then, the progress position is marked on the text information together with the topic keyword. For example, the timing of the start and end of a comment is marked. The minutes information database 15 records sentence pattern data including “topic keywords” and “feature words” that guide the agenda.
[0040]
For example, if "start", "speak", and "next" are registered as "feature words", it is assumed that the following text information is obtained.
(1) "Begin" the current R & D discussion.
(2) “Next,” I will talk about future research and development.
(3) “Next”, “Discussion” on research expenses will begin.
At this time, “current research and development”, “future research and development”, and “research expenses” are obtained as “topic keywords” corresponding to “characteristic words”.
[0041]
The marked text information extracted by the voice recognition processing unit 14 is stored in the minutes information database 15.
[0042]
The terminal can request the meeting record in which only the key points based on the “topic keyword” are recorded from the voice communication recording server 1. At this time, the terminal transmits a conference minutes request message including the “topic keyword” to the voice communication recording server 1.
[0043]
The key point search unit 16 acquires the conference record request message received by the communication interface 19. Then, the gist search unit 16 searches the minutes information database 15 based on the “topic keywords”. For example, if the requested "topic keyword" is "future research and development", a mark corresponding to "future research and development" is searched from the minutes information database 15, and the progress position such as start and end of the mark is searched. Information can be obtained.
[0044]
The minutes record request message is not limited to “topic keywords”, “participant IDs (for each speaker)” and “speech time”, but may be “free text information” or “free speech information”. It may be. “Free text information” refers to text information of a sentence that describes the content that the meeting requester wants to obtain. The voice communication recording server 1 can extract topic keywords and the like from the text information. The “free voice information” refers to voice information in which the contents requested by the conference record requester are recorded. The voice communication recording server 1 can extract text information by performing voice recognition processing on the voice information, and can extract a topic keyword or the like from the text information.
[0045]
The gist search unit 16 can also refer to the synonym database 18. For example, the feature word "begin" is synonymous with "begin,""open," and "do." These pieces of information are systematically configured and stored in the synonym database 18. Therefore, the key point search unit 16 can also search the minutes information database 15 for synonyms of the characteristic words. Of course, it is preferable that the synonym database 18 can also search for synonyms of topic keywords.
[0046]
The meeting record generation unit 17 searches the audio information database 10 and the video information database 11 based on the mark position and the time information based on the text information searched by the gist search unit 16. The conference record generation unit 17 acquires audio information and video information corresponding to time information added to the searched text information. Then, it generates an AV (Audio and Visual) meeting record in HTML format, and returns the meeting record to the terminal that transmitted the meeting request message.
[0047]
Thus, the terminal that has transmitted the conference request message can receive only the conference based on the requested “topic keyword”, and the terminal can reproduce the audio and video in the conference situation.
[0048]
FIG. 3 is a sequence diagram between the voice communication recording server and the terminal according to the present invention.
[0049]
According to FIG. 3, it proceeds in the following sequence.
(S30) When starting a remote communication conference, a certain terminal transmits a conference start registration message to the voice communication recording server 1. This conference start registration message includes the “conference name”, “participant name”, “start time”, and the like of the conference to be started. The audio communication recording server 1 determines the conference ID and the participant ID from these pieces of information by the conference ID determination unit 13.
(S31) The voice communication recording server 1 distributes the conference ID and the participant ID to the terminals 2, 3, and 4 of the participants.
(S32) First, each of the terminals 2, 3 and 4 performs time synchronization with the voice communication recording server 1. Note that each of the terminal and the voice communication recording server 1 may access the NTP (Network Time Protocol) server or the SNTP (Simple Network Time Protocol) server on the Internet, and may match the built-in clocks.
(S33) The terminals 2, 3 and 4 start a conference via the Internet 5. The audio information in this conference is obtained from the microphone of each terminal, and the video information is obtained by the video camera of each terminal. The audio information and the video information to which the conference ID and the participant ID have been assigned are distributed to the audio communication recording server 1.
(S34) When ending the telecommunication conference, a certain terminal transmits a conference end registration message to the voice communication recording server 1.
(S35) Thereafter, a participant may need a minutes based on a desired “topic keyword”. At this time, the terminal transmits a conference request message to the voice communication recording server 1 by the operation of the participant.
It should be noted that the terminal transmitting the conference record request message is not necessarily limited to a terminal of a participant of the conference (a terminal transmitting audio information or the like).
(S36) Based on the “keyword” included in the conference record request message, the audio communication recording server 1 creates an HTML format AV conference record combining audio information and video information. Then, the minutes are transmitted to the terminal that transmitted the minutes request message.
[0050]
FIG. 4 is a flowchart when the audio communication recording server according to the present invention receives audio information and video information of a conference.
[0051]
According to FIG. 4, it proceeds in the following sequence.
(S40) A conference start registration message is received from the terminal.
(S41) Based on information such as "conference name" and "participant name" included in the conference start registration message, a "conference ID" and "participant ID" are determined, and these IDs are distributed to the participants' terminals. I do.
(S42) Audio information and video information during the conference are received from the terminal.
(S43) The current time is stamped on the audio information and the video information. This is effective for the streaming format. On the other hand, in the case of the file format, if the time is stamped on the file at the terminal, there is no need to stamp the time here.
(S44) The received audio / video information is distinguished between audio information and video information.
(S45) The voice information is stored in the voice information database 10 for each participant.
(S46) The video information is stored in the video information database 11 for each participant.
(S47) The voice information is converted into text information by voice recognition processing. At this time, for the text information, a feature word meaning the progress position of the topic of the meeting is searched, a topic keyword corresponding to the feature word is searched, and the progress position is marked on the text information together with the topic keyword. At this time, a synonym database 18 in which synonyms for the characteristic words are stored in advance is used. By using this, a synonym for the characteristic word is searched, and a topic keyword corresponding to the synonym is searched.
(S48) The text information is stored in the minutes information database 15 for each participant.
[0052]
FIG. 5 is a flowchart when the voice communication recording server according to the present invention receives a conference request message.
[0053]
According to FIG. 5, the process proceeds in the following sequence.
(S51) A conference record request message is received from the terminal. The conference minutes request message includes a “topic keyword”.
(S52) The minutes information database 15 is searched for text information based on "topic keywords".
(S53) The time information of the searched text information is specified.
(S54) The audio information and the video information corresponding to the specified time information are acquired from the audio information database 10 and the video information database 11.
(S55) A conference record including the acquired audio information and video information is generated. Here, the minutes are in HTML format. As a result, multimedia minutes can be provided.
(S56) The generated minutes are transmitted to the terminal that transmitted the minutes request message.
[0054]
Each step in the voice communication recording / reproducing method of the present invention can be executed by using a recording medium built in a computer and using control means such as a CPU. Also, a computer-readable program can be installed via a recording medium such as a CD or a communication line and executed by the computer. These programs may be mainly realized as a function of a server on the Internet by a program mounted on the server. Of course, these functions are also realized by a program installed in the terminal, and can be used in a peer-to-peer type.
[0055]
【The invention's effect】
As described above in detail, according to the voice communication record generation method, apparatus, and program of the present invention, when providing a conference record, conditions desired by a person who needs the conference record (topic keyword, speaker, etc.) ) Can provide the minutes of the meeting. In particular, in a telecommunication conference using the Internet, the creation of minutes in the present invention adds value to existing services.
[0056]
As a result, after the meeting, the specific speaker who wants to hear again (for example, the statement of the president), the content of the discussion of the head of the meeting that was missed because it was late, or only the summary of each agenda as a summary of the meeting Reviews can be obtained. In particular, it is possible to acquire a compact conference record of only a necessary portion corresponding to a desired condition without acquiring the entire conference.
[Brief description of the drawings]
FIG. 1 is a system configuration diagram according to the present invention.
FIG. 2 is a functional configuration diagram of a voice communication recording server according to the present invention.
FIG. 3 is a sequence diagram between a voice communication recording server and a terminal according to the present invention.
FIG. 4 is a flowchart when the audio communication recording server according to the present invention receives audio information and video information of a conference.
FIG. 5 is a flowchart when the voice communication recording server according to the present invention receives a conference record request message.
[Explanation of symbols]
Reference Signs List 1 voice communication recording server, voice communication recording device 10 voice information database, voice storage means 11 video information etc. database 12 time stamp section 13 meeting ID determination section 14 voice recognition processing section 15 minutes information database, storage means 16 requirement search section 17 Conference record generation unit 18 Synonym database, synonym storage means 19 Communication interface units 2, 3, 4 Terminals 21, 31, 41 Microphones 22, 32, 42 Video camera 5 Internet

Claims (12)

A step of converting the voice information into text information by voice recognition from voice information from at least two or more points and voice storage means for storing the voice information together with the input time information;
Determining time information in which the text information includes at least one topic keyword or a synonym thereof from the topic keyword storage unit that stores the topic keyword, and recording the time information and the topic keyword in the recording unit in association with each other; ,
A step of determining time information at which a topic keyword or a synonym thereof is recorded in the recording unit in the input search request information;
Outputting said determined time information or voice information corresponding to a predetermined time section including said time information from said voice storage means. The method according to claim 1, further comprising a step of detecting a speech section and a silent section of the voice information. 3. The voice communication record generation method according to claim 1, wherein a synonym is further searched for using a synonym storage unit that previously stores synonyms for the topic keyword, and the progress position is marked on the text information together with the topic keyword. 4. The audio communication record generation method according to claim 1, wherein conference information is added to the audio information, and other information is linked to the audio information. Means for converting the voice information into text information by voice recognition from voice information from at least two or more points and voice storage means for storing the voice information together with the input time information;
Means for determining time information in which the text information includes at least one topic keyword or a synonym thereof from the topic keyword storage means for accumulating topic keywords, and recording the time information and the topic keyword in the recording means in association with each other. ,
Means for determining time information at which a topic keyword or a synonym thereof is recorded in the recording means of the input search request information;
A means for outputting the determined time information or voice information corresponding to a predetermined time section including the time information from the voice storage means. The voice communication record generation device according to claim 5, further comprising means for detecting an utterance section and a silent section of the voice information. 7. The voice communication record generation device according to claim 5, wherein a synonym is further searched for using a synonym storage unit that previously stores synonyms for the topic keyword, and the progress position is marked on the text information together with the topic keyword. 8. The voice communication record generation device according to claim 5, wherein conference specific information is added to the voice information, and other information is linked to the voice information. 9. A step of converting the voice information into text information by voice recognition from voice information from at least two or more points and voice storage means for storing the voice information together with the input time information;
Determining time information in which the text information includes at least one topic keyword or a synonym thereof from the topic keyword storage unit that stores the topic keyword, and recording the time information and the topic keyword in the recording unit in association with each other; ,
A step of determining time information at which a topic keyword or a synonym thereof is recorded in the recording unit in the input search request information;
A voice communication record generation program for causing a computer to execute the process of outputting the determined time information or voice information corresponding to a predetermined time section including the time information from the voice storage unit. The voice communication record generation program according to claim 9, wherein the computer is configured to execute a computer having a step of detecting an utterance section and a silent section of the voice information. 11. The computer according to claim 9, wherein a synonym is further searched for using a synonym storage unit that previously stores synonyms for the topic keyword, and a computer is executed so as to mark the text information together with the topic keyword in the progress information. 12. Voice communication record generation program. The conference specific information is added to the audio information, and a computer is executed so that other information is linked to the audio information. Voice communication record generation program.

Images