JPS607528A - Voice-controlled word processor - Google Patents

Abstract

PURPOSE:To increase the input speed of a voice-controlled word processor by converting the input of said word processor into a more natural voice input and at the same time decreasing the number of operation procedures of KANJI (Chinese character) conversion. CONSTITUTION:The input of a voice-controlled word processor is converted into a more natural voice input, and at the same time the number of operation procedures for KANJI conversion is reduced. For instance, the voices of an operator are supplied to a voice recognizer 11 through a mike 12. Then the result of recognition is coded and sent to a processor 13, and the processor 13 converts the result of recognition into the corresponding characters to display them in a display device 16. When the KANA (Japanese syllabary) fed by the operator is converted into KANJI, a conversion key on a keyboard 17 is depressed. Thus the candidate KANJI corresponding to the KANA fed from a KANJI dictionary 14 via processor 13 is selected and displayed in a display device 16. Then the character string fed finally is stored in a document memory 15 and then printed by a printer 18 with push of a print key on the keyboard 17.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a voice word processor that inputs characters by voice on a keyboard and inputs, proofreads, edits and prints sentences.

Conventionally, in this type of spoken word processor, the operator converted the words into ``a'', ``i'', ``u'', and ``e'' character codes. This character code is stored in the process.

In this way, in a word processor that inputs monosyllables, that is, kana characters, the speech recognition device must recognize the input monosyllabic speech. However, since monosyllabic speech contains little speech information, it is easy to be misrecognized, and in this case, the operator has to utter the same speech many times until it is correctly recognized. This drawback is particularly noticeable in the characters ``<<'' and ``n'', which have little vocal energy when uttered as a single sound. In addition, since the operator also has to input the text by dividing it into monosyllables, it is difficult to input, and it takes a long time to input.

Furthermore, when converting such monosyllabic kana input into kanji, the operator first presses the ``kanji key'' on the word processor's keyboard to inform the processing unit W that the kanji conversion is to be performed, and then converts the voice into kanji. After inputting the kana that corresponds to the desired kanji, the user presses the ``conversion key'' to instruct the kanji conversion of the kana input after pressing the ``kanji key.'' Therefore, the procedure for converting kana to kanji is complicated, resulting in a decrease in character input speed. Therefore, it is an object of the present invention to further increase the input speed of a spoken word processor by making the input to the spoken word processor more natural and reducing the number of operating procedures for kanji conversion.

According to the present invention, there is provided a speech recognition device that registers the reading of a polysyllabic Kanji as a recognized speech and converts the input speech into a recognition code and outputs it, and a Kanji dictionary that accesses the corresponding Kanji using a Kana character code. and a processing device that converts recognition codes from a speech recognition device into kana character codes, accesses a kanji dictionary, and obtains kanji characters corresponding to the kana character codes.

According to research by various organizations, if you prepare about 2,000 character types for use in Japanese documents, you can express more than 90 characters in Japanese documents, but even if you increase the number of character types even more. The rate of Japanese sentences that can express
It is reported that he does not approach 00chi. Therefore, if you prepare common or regular kanji as kanji, there will be almost no problem in creating Japanese sentences.

Furthermore, we found that many kanji belonging to the kanji category have the same reading as shown in Figure 1, and in the case of on-yomi,
I discovered that the 100 readings of Habo can cover about 80 sections of kanji for general use. Therefore, when registering speech patterns in a speech recognition device, instead of only registering monosyllabic sounds as in the past, readings of kanji such as "in", "ei", "eki", and "sho" are also registered as words. If you do this, you will be able to directly access the kanji using this recognition result. By registering ``Kuyou'', the operator's multi-syllabic voice ``Shou'' can be directly converted into Kana, resulting in 56 different types such as ``小'', ``masu'', ``noboru'', etc. You will be able to access useful kanji.

Next, the present invention will be described in detail with reference to all the drawings showing one embodiment of the present invention.

In Figure 2, the voice uttered by the operator is transmitted through microphone 1.
2 is input to the speech recognition device 11. As a result, the recognition result of the recognition device 11 is encoded into the processing device 1.
Sent to 2. The processing device 12 has an internal table for converting the code given from the recognition device 11 into a kana character code, and converts the recognition result into the corresponding character.
The information is displayed on the display device 16. The type of displayed character for each character (cal, alphabetic character, etc.) is specified by the operator using a group of input auxiliary keys on the keyboard 17 before voice input, but normally the mode is set to kana mode. If the operator wishes to convert the input kana into kanji, the operator presses the "conversion key" on the keyboard 17, and the processing device 13 accesses the kanji dictionary 14 and converts the input and displayed kana into kanji. Candidate kanji are selected and displayed in order on the display device 16 according to the operator's instructions, that is, by pressing the "conversion key" again. The finally input character string is stored in the document storage device 15, and when printing it on paper, the operator uses the keyboard 17.
By pressing the print key above, the output is output to the printer 18 and printed.0 The above is the general configuration and operation of the voice word processor. Anne', 'in'.

Approximately 100 types of polysyllabic sounds such as "ei", etc. are registered. In other words, the operator can input multisyllables into these registered voices instead of inputting monosyllabic voices.

3. How to register voice to this voice recognition device 11
This will be explained with reference to the figures.

The memory area for storing memory codes of speech recognition device #11 is in a matrix form, and coordinates called segment and word numbers are assigned.

The number of segments and the number of words are set arbitrarily in advance, but the number of words is determined by using the upper letters shown in Figure 2, alphabetical characters in segment 2, numbers in segment 3, and symbols in segment 4. , the segment 5 operator first presses the "Voice Registration" key on the keyboard 17 in order to register his/her own voice to the voice recognition device 11VC. Processing device 13
gets this. When the speech recognition device 11 is able to perform this quasi-11i1i, it returns a ready response to the processing device 13. As a result, the processing device 13 causes the display device 16 to display the character "a" to be registered in the area specified by segment 1 and word 1. In response to this display, the operator utters "a", which is processed by the speech recognition device 11, becomes a voiceprint, and is registered in the area designated by segment 1 and word 1. Next, the processing device 13 instructs the recognition device 13 to register "i" in the area specified by segment 1 and word number 2.
instruct. Thereafter, the voices are registered in the same way, but segment 5 includes "An", "In", and "En".

The sounds of kanji such as ... are registered one after another. By repeating this process, all the voices that can be recognized are stored in the memory area for storing the recognition code of the recognition device 11.

Once the voice registration has been completed, the voice recognition device 11 converts the voice input from the operator into an electrical signal and compares it with the voice stored in the recognition code table to find a match or the most similar area segment. and the word number as the recognition result code. For example, if the operator [A
If NJ is uttered, segment 5 and word number 1 are output to the processing device 13 as the recognition result code. Upon receiving this, the processing device 13 converts fL into a kana character string "an".

The method of this conversion within the processing device 13 will be explained with reference to FIG.

A table is provided in the processing device 13 for converting the segment numbers and word numbers sent from the speech recognition device 11 into kana codes. In this table, the upper 4 bits of the address are created from the segment number, and the lower 8 bits of the address are created from the word number. The segment number is directly converted to hexadecimal and the upper 4 of the address
The word number is multiplied by 4 and converted to hexadecimal to form the lower 8 bits of the address. for example,
Assuming that segment 5 and word number 1 are input, the processing unit 13 converts these numbers into binary numbers 010100000.
The table is accessed by converting the address to 001, that is, 0501 in hexadecimal (however, the first 0 is a dummy number). In other words, every fourth address in the table corresponds to consecutive segment numbers and word numbers. In this way, one address is accessed by one segment number and word number, but the stored contents of the three addresses following the designated address are also read from the table. Since a 1-character Kana code is stored in the area specified by 1 address in the table, a maximum of 4 characters can be accessed from the table by accessing 1 address created from 1 segment number and word number. The code will be read.

In FIG. 4, for example, the area designated by address 0504 stores a 1-byte kana code for "A", and the area designated by address 0505 stores a 1-byte kana code for "n". Addresses 0506 and 0507 are empty areas. When the recognition result code of segment 5 and word number 1 is input from the speech recognition device 11, the processing device 13 converts this code to address 0504 and accesses the table using this address. This results in addresses 0504, 0505, 0506. and 0507
The contents of ``a'' and ``n'', that is, the two-character kana code, are read out.

The processing device 11 can access the Kanji dictionary 14 using the read Kana code and obtain the corresponding 2-byte Kanji code.

In this way, since voices are registered in the voice recognition device 11 in units of phonetic pronunciation of kanji, the operator can obtain one kanji with one continuous pronunciation.

As described above, the present invention has the effect of reducing the operator's operations for inputting Chinese characters and improving the input speed by registering the readings of multi-syllable Chinese characters in the speech recognition device It.

[Brief explanation of drawings]

Figure 1 shows some of the regular kanji that have the same reading.
FIG. 2 is a block diagram of an embodiment of the present invention, FIG. 3 is a diagram showing a voice registration area of a speech recognition device, and FIG. 4 is a diagram showing a kana code storage area of a processing device. Figure 7 Pore No. 1 234 Yu- degu 2 ABCD Me: ni 6 Thai Gugu Tazukun 3 Figure

Claims (1)

[Claims] a speech recognition device that registers the reading of multi-syllable kanji as recognized speech and converts the input speech into a recognition code and outputs it; a kanji dictionary that accesses the corresponding kanji using a kana character code; A voice word processor comprising: a processor for converting into kana character codes and accessing the kanji dictionary.