JPH02206825A - Device for preparing minutes - Google Patents

Abstract

PURPOSE:To securely and speedily prepare minutes without depending on a manual operation and to correct the content of a statement while the processing is expedited by automatically making a sentence from the voice of the speaker of the processing, and preparing a summarized sentence so as to visualize it. CONSTITUTION:In a voice recognition device 4, a voice signal from a microphone 1 and the like is segmented at every constant period, and a characteristic parameter is extracted by Fourier-transform and the like so as to segment words. The device 4 segments the words and identifies the speaker of the voice signal concerned by comparing the words with the voice characteristic parameter of an attendance registered in an attendance voice registration device 3. A sentence preparation device 5 fetches a word string which has been speaker- identified from the device 4 together with speaker identification information, analyzes a construction and a meaning by considering the context in terms of words. Furthermore, the device 5 prepares the meaningful sentence by bringing in a dictionary for processing and the like. A sentence summary device 6 fetches information from the device 5, which to made into the sentence, summarizes it, and displays it in a display device 8.

Description

[Detailed description of the invention] A. Industrial application field The present invention relates to a device for creating minutes.

B1 Overview of the Invention The present invention automatically recognizes the speaker's voice, converts it into text, summarizes the text, and visualizes it, thereby reliably and quickly creating the minutes without using any human hands. It has been made possible to do so.

C9 When creating the minutes of conventional technical meetings, lectures, training sessions, etc., one of the attendees should act as the recorder and prepare the minutes professionally.
Audio is recorded using a tape recorder, etc., and the person in charge later listens to it again to create minutes.

The problem that the D0 invention aims to solve In the former method of creating minutes, it becomes difficult for the recorder to participate in the proceedings, and a specialized stenographer is often required. On the other hand, in the latter preparation method, the preparation of the minutes is delayed by 4 hours, so there is a considerable delay in re-checking the speaker for corrections and error corrections after the minutes have been prepared.

An object of the present invention is to provide a minutes creation device that can automatically create minutes reliably and quickly without using any human labor.

Means for Solving 80 Problems In order to achieve the above object, the present invention provides an attendee voice registration device for registering voice characteristic parameters of conference attendees;
a speech sound input device that obtains a speech signal of a speaker, a speech recognition device that cuts the speech signal into words based on its feature parameters and identifies the speaker by comparing the speech feature parameters with the speech feature parameters; a sentence creation device that converts speech feature parameters based on spoken words and speaker identification information into sentences by syntactic analysis and semantic analysis; a sentence summarization device that summarizes sentences from the transcribed data; It is equipped with a minutes output device that records sentences together with the name of the speaking number, recognizes the voice of the speaker by speech sound, converts it into sentences, further summarizes and visualizes the sentences, and automatically performs these processes.

F. Embodiment FIG. 1 is a block diagram showing an embodiment of the present invention. The microphone I is used as a speaker input means for inputting the speech voice of the speaker, and may be replaced with a voice recorded by a tape recorder 2 or the like.

Prior to the start of the proceedings, the attendee voice registration device 3 registers the voice characteristic parameters of each attendee as table data by inputting the characteristic voices of all the attendees and their name pronunciation sounds.

The speech recognition device 4 cuts out the speech signal from the microphone 1 or the like at regular intervals (for example, 5 to 10 m5ec), extracts characteristic parameters by Fourier transform, etc., and cuts out words. Word extraction from this speech feature parameter is
For example, this can be realized by the method described in the biweekly magazine, Nikkei Electronics Publishing, ``Applying neural net to voice typewriter'' June 13, J19B8 issue.

The above-mentioned speech recognition device 4 performs word extraction, and also performs speaker identification of the speech signal by comparison with the attendee speech feature parameters registered in the attendee speech registration device 3.

The sentence creation device 5 takes in the speaker-identified word string together with the speaker identification information from the speech recognition device 4, performs syntactic analysis and semantic analysis taking into account the context of each word, and further refers to a dictionary of proceedings terms and the like. generate meaningful sentences.

The text summarization device 6 takes in text information from the text creation device 5 and summarizes the information. This text summary will be explained with reference to FIG.

First, the keyword extraction unit 11 automatically extracts keywords from the input source text and analyzes the nature of the content of the statement. This analysis involves analyzing whether the sentence in question is a question, an answer, a meaning, or an impression, and then analyzing whether the answer or meaning is positive or negative. is registered in the comment attribute table 12.

Next, the table! The statement attribute data of No. 2 and the keywords of the keyword extraction unit 11 are registered together with the text in the statement contents table 13 and the proceedings progress table 14 for each speech sound, using the speaker and the topic as keys. Furthermore, table 1
After analyzing the content of 4, the number of luxurious opinions, 9 dissenting opinions, and their contents were statistically analyzed, and the results of this analysis are the minutes progress statistics table! Registered in 5.

The data in each of the tables 13, 14, and 15 described above is transferred to the sentence summary unit 16 using neuromemory, and the summary unit!
In step 6, an associative operation is executed by referring to the keyword, the attribute data of the comment attribute table I2, the synonym/antonym dictionary section 17, and the summary dictionary 18, and a summary sentence is created.

Such a text summary can be created, for example, when the proceedings are uttered and written by speakers A, B, and C according to the following procedure: (A1) I'm busy today...I'd like to start a meeting about new product X. .

(B1) The development plan for new product X is as follows.

(CI) Are you able to keep to your schedule on time? (B2
)... is a problem, but it is expected to proceed as planned.

For the above-mentioned amount of original text, keyword extraction is new product Z, development plan, schedule, time, etc., and comment attributes are agenda presentation for comment (AI), introduction, comment (for comment (Bl)), etc.
CI) is a question, and statement (B2) is an answer. Also,
In the statement (C1), the question number table data is incremented by 1.

An example of the summary results from such an analysis is as follows.

(AI) New product X meeting.

(Bl) Introduction of new product development plan.

(CI) Development schedule question.

(B2) The plan is generally on schedule.

Next, returning to FIG. 1, the sentence summarized by the sentence summarizing device 6 is printed and displayed on the kanji printer 7 and the display device 8 together with the speaker's name. Furthermore, when the sentence summarization device 6 cannot create a comment attribute table or determine whether a comment attribute is affirmative or negative, the speech synthesis device 9 and the audio output bag M10 are used as a means of transmitting the non-summarization to output a speech indicating that the text cannot be summarized. . For example, when you cannot summarize, you can register it as a fixed phrase. [The current statement cannot be summarized.] Please tell me the purpose again. ] The text is synthesized into speech, and the audio output is obtained from a speaker or the like. This unsummarizable communication helps correct and clarify ambiguous statements.

Based on the above, in this embodiment, voice recognition, transcription, summarization, and visualization are automatically performed every time a statement is made by all participants in the discussion, and if the content of the statement is unclear, the speaker is prompted to repeat the statement. Increasing the density of the proceedings by ensuring the recognition and writing, etc., and ensuring the content of what is being said. Then, through automation, the minutes can be created almost simultaneously with the statements being made, eliminating the need for manual minutes.

Effects of the G0 Invention As described above, according to the present invention, since the voice of the speaker is automatically converted into text and a summary sentence is created and visualized, the discussion can be carried out reliably and quickly without relying on human hands. Recordings can be made, and comments and corrections can be made during the proceedings.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the present invention, and FIG. 2 is a block diagram of the text summarization device shown in FIG. l...Microphone, 3...Attendee voice registration device,
4...Speech recognition device, 5...Text creation device, 6...
- Sentence summary device, 7... Kanji printer, 8... Display device, 9... Speech synthesis device, 11... Keyword extraction unit, 12... Speech attribute table, 13...
・Remarks content table, 14... Minute progress table, 15... Minute progress statistics table, 16... Sentence summary section, 17... Synonyms/contras! ! Word dictionary, 18... summary side dictionary.

Claims (1)

[Claims] (1) An attendee voice registration device that registers the voice characteristic parameters of the conference attendees, a voice input device that obtains the voice signal of the voice speaker, and a voice input device that extracts the voice signal from the voice characteristic parameters into words. and a speech recognition device that identifies a speaker by comparing the speech feature parameters with the speech feature parameters;
a sentence creation device that converts the extracted words and speech feature parameters based on the speaker identification information into sentences by syntactic analysis and semantic analysis; a sentence summarization device that summarizes sentences from the text-formed data; and the summarization device. What is claimed is: 1. A minutes creation device comprising: a minutes output device for recording written sentences together with names of speakers.