JPS6281699A - Dictionary creation and updating method in audio word processing device - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a dictionary creation and updating method in an audio word processing device.

Conventional Vampire In conventional speech word processing devices, the user must register all the necessary words, which requires a great deal of effort.

Furthermore, each time a misrecognition is made, the dictionary must be recreated, and even when performing an update process, the user is required to input correct and incorrect answer information. Additionally, there are some tools that perform update processing on existing dictionaries, but the recognition rate can be improved more by replacing the user's dictionary rather than updating it.

The present invention was made in view of the above-mentioned circumstances.
In particular, this was done with the aim of efficiently performing dictionary creation and updating operations in an audio word processing device.

1. In order to achieve the above object, the present invention has means for exchanging input with a registered dictionary, means for further updating the dictionary, and means for modifying displayed recognition results using a keyboard. A voice word processing device that recognizes a registered monosyllable or word by saying it, performs kana-kanji conversion and Japanese processing, and displays the correct sentence. The feature is that recognition is performed using a dictionary, and once the incorrectly recognized part is corrected by the user, the voice data input for the first word selected is replaced with that of the dictionary, and the dictionary is gradually specified. That is. Hereinafter, an explanation will be given based on one embodiment of the present invention.

When creating a dictionary for an audio word processing device, the number of words is large, and especially for an audio word processing device that automatically distinguishes between about 1,000 words and two monosyllables, the registration work requires a lot of effort. . The present invention has been made in view of the above-mentioned circumstances, and is particularly designed for users who use it for the first time to prepare a word dictionary in advance and use it (each time, the dictionary is made specific to the user). It aims to eliminate the initial registration effort.

FIG. 1 is an electrical Bullots diagram for explaining one embodiment of the present invention, in which 1 is a microphone, 2 is a feature extraction unit, and 3
4 is a control unit, 4 is a monosyllable recognition unit, 5 is a word recognition unit, 6 is a result processing unit, and 7 is a result display unit.

8 is a user, 9 is a correction section, 10 is a word dictionary section, and below,
An embodiment of the present invention will be described using the BTSP method. Note that audio word processing devices may be capable of only using words or only single syllables, but here we will use approximately 1,000 words and 100 monosyllables.
Let's consider something that has something like a word and has a function to automatically sort it. Further, it is assumed that in the word processing device according to the present invention, words are fixed in advance. Therefore, when using the above-mentioned speech word processing device, registering all the above-mentioned words requires a great deal of effort as mentioned above. Therefore, first prepare a general dictionary. For example, when using the BTSP method, this general dictionary uses a weighted average dictionary that overlaps the speech patterns of several people with the same value to absorb the unspecificity to a certain extent, and the were selected from people with average vocalizations.

Of course, the recognition rate is worse than when the speaker is identified;
It depends on the person, but it should be around 80%. A user first installs this dictionary in a word processing device and causes recognition to be performed on this dictionary. For example, when a sentence as shown in FIG. 2(a) is input, it is assumed that the first dictionary recognizes it as shown in FIG. 2(b). In this case, the user can use the keyboard to
At the point when the correction is completed in the L' part (C), starvation, sight, and joten.

-na 0. - Obtain Lee's dictionary (the above-mentioned weighted average dictionary), replace all the input voice data with the existing dictionary, and use it as the dictionary from now on. When a word that is uttered for the first time is corrected, it is replaced with the dictionary and the word is specified to the user.

Next, if a similar sentence is input and it is misrecognized as in (d), it will be corrected again, but this time the user's dictionary has already been registered, so the update process will be performed when the user corrects it. be exposed. Also, when the next sentence (e) is input, ■ change the dictionary for heaven only. In this way, for a word or monosyllable input for the first time, the input data and the correct recognized word are replaced when the user completes the correction, and the update process is performed only when the word is selected for the second time or later, or when it is misrecognized. , and gradually specialize the dictionary. Naturally, by saving this, the user will hardly need to perform any registration operations thereafter. In this way, after a certain amount of use, it becomes a specific dictionary for that user.

As is clear from the above description, according to the present invention, there is no need to register words necessary for the audio word processing device, and furthermore, it can be used as a high-quality specific dictionary each time it is used.

[Brief explanation of drawings]

FIG. 1 is an electrical block diagram for explaining an embodiment of the present invention, and FIG. 2 is a diagram for explaining an example of a dictionary creation and updating method according to the present invention. DESCRIPTION OF SYMBOLS 1... Microphone, 2... Feature extraction part, 3... Control part, 4... Monosyllable recognition part, 5... Word recognition part, 6...
・Result processing section. 7...Result display section, 8...User 29...Correction section, 10...Word dictionary section.

Claims (2)

[Claims] (1) It has a means for exchanging input with a registered dictionary, a means for updating the dictionary, and a means for modifying the displayed recognition result using a keyboard, and utters a pre-registered monosyllable or word. For first-time users, a voice word processing device that recognizes the word, performs kana-kanji conversion and Japanese processing, and displays the correct sentence. Dictionary creation and updating in a voice word processing device, characterized in that the dictionary is gradually specified by replacing the voice data input for a word selected for the first time when the misrecognized part is corrected, and by gradually specifying the dictionary. Method. (2) A method for creating and updating a dictionary in an audio word processing device, characterized in that when a dictionary that has already been replaced is incorrectly recognized, the dictionary is updated at the time of correction.