JPS61239358A - Documentation system by voice input - Google Patents

Abstract

PURPOSE:To easily prepare a conversation record by causing a word processor to automatically input the name of the speaker of each speaking part. CONSTITUTION:Inputted voices are separated into syllables by a speech recognizing device 2 and supplied to a comparator 8 one syllable at every syllable. At the beginning of each speaking part, the comparator 8 compares the voice pattern sent from the device 2 with the voice patterns of all speakers registered in a voice pattern registering table 10 and converts the voices into corresponding characters. Moreover, the registered of the voice pattern having the highest matching ratio to the voice pattern of the inputted voices is discriminated as the speaker of the speaking part and the character data of the name of the relevant speaker are read out from the table 10. After the speaker is determined, inputted voices are compared with the voice pattern of the relevant speaker only and converted into characters. The character data and speaker name are supplied to a CPU 1 from the speech recognizing device 2 and stored in a buffer 7 through the CPU 1 and at the same time, displayed on a display device 3.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a document creation method in which a document is created by inputting dialogues of a plurality of speakers into a word processor by voice [Prior Art] A word processor using a voice input method is , which has the advantage of greatly simplifying input operations. In this type of word processor, the voice of a document creator is registered in advance, and when voice input is actually performed, the voice is identified based on the registered voice.

By the way, it is possible to create minutes of a conversation or meeting in which multiple speakers appear using a voice input type word processor if the voices of each speaker are registered in advance. However, in a conversation, etc., the speaker does not say his or her own name each time he or she starts speaking, so if it is necessary to add the speaker's name when creating a document, the operator must write the current speaker's name separately. must be entered.

[Structure of the Invention] The present invention aims to make it possible to easily input the names of speakers when creating a document by inputting dialogues between multiple speakers into a word processor by voice. In this method, the speaker name and voice pattern of each speaker are registered in the word processor in advance, the input voice is compared with the voice pattern of each speaker to determine the speaker, and the speaker name is recorded along with the voice. It is characterized by automatic input.

[Effects of the Invention] As described above, in the present invention, since the word processor automatically inputs the speaker's name for each conversation part, it becomes possible to easily create this type of dialogue frame. Furthermore, since the word processor itself inputs the speaker's name, dialogue can be input directly into the word processor and edited in real time.

[Embodiment] An example of a voice input type word processor employing the method of the present invention will be described below.

As shown in the drawing, the voice input word processor according to the present invention includes a CPU (central processing unit) that performs editing control, a voice recognition device 2 that recognizes input voice, and a display device 3 (CRT display, etc.). , a keyboard 4, a floppy disk 5, and other external storage devices that can be connected to and separated from the word processor main body, a printer 6, and a buffer 7 that stores sentences input by voice input or key input.

The voice input device 2 is connected to the voice pattern registration table IO via a comparison device 8.

This voice pattern registration table IO has a plurality of registration areas 11 each consisting of a name area Ila and a voice area Ilb. In each name area lla, the name of each speaker appearing in the dialogue to be recorded is stored in advance, while in each voice area Ilb, the voice pattern of each speaker is stored in correspondence with each name area lla. Stored in advance.

When creating a dialogue record, the dialogue by the speakers registered in the voice pattern registration table 1O is input directly to the voice recognition device 2 from the microphone 12, or the dialogue recorded on the cassette tape 13 is input to the voice recognition device 2. The dialogue by the person is input indirectly.

The input speech is separated into syllables (for example, I characters such as ah, ka, . . . ) by the speech recognition device 2, and is supplied to the comparison device 8 syllable by syllable.

At the beginning of each conversation part, the comparison device 8 compares the voice pattern sent from the voice recognition device 2 with the voice patterns of all speakers registered in the voice pattern registration table IO, and converts it into corresponding characters. do. In addition, at that time, the voice pattern of the input voice and the one with the highest matching rate (
The registered bundle of voice patterns (best matching) is determined to be the speaker of the conversation part, and character data of the speaker's name is read from the voice pattern registration table IO.

After the speaker is determined, the input voice is compared only with the voice pattern of the speaker and converted into text.

In this way, the speaker name and character data corresponding to the input voice are supplied from the speech recognition device 2 to the CPUI, stored in the buffer 7 via the CPUI, and displayed on the display device 3. Ru.

In that case, each part of the conversation is enclosed in parentheses, and the speaker's name is added before the first parenthesis (Yamada ``×××...
.... ).

Note that the end of each conversation part is determined as follows.

In other words, if the input audio is interrupted for a certain period of time or more, the input audio is changed to the audio pattern registration table 1 again.
The speech patterns of all speakers within 0 are compared. If the newly determined speaker is the same as the previous speaker, it is assumed that the conversation has not yet ended.

On the other hand, if the newly determined speaker is different from the previous speaker, it is assumed that the previous conversation has ended and an end parenthesis is added. Next, add a line break, treat the interrupted conversation as a new conversation, and add a new speaker's name.
It is stored in the buffer 7 and displayed on the display device 3.

By repeating these steps, a dialogue record is created.

After making necessary corrections using the keyboard 4, the created dialogue record can be printed using the printer 6.

Note that the character data accumulated in the buffer 7 can be transferred to the floppy disk 5 and stored on the floppy disk 5, if necessary.

[Brief explanation of the drawing]

The drawing is a block diagram of a voice input type word processor employing the method of the present invention. 10...Voice pattern registration table.

Claims (1)

[Claims] (1) When creating a document by inputting the dialogues of multiple speakers into a word processor by voice, register the speaker name and voice pattern of each speaker in the word processor in advance, and input the input voices to each A document creation method using voice input, characterized in that the speaker is determined by comparing the speaker's voice pattern, and the speaker's name is automatically input along with the voice.