JPS6146557A - Speech word processor - Google Patents

Abstract

PURPOSE:To improve the input speed by providing a code convension table and a Kanji (Chinese character) dictionary memory in a processor to automatize the retrieval operation of candidate characters resulting from voice recognition in a certain degree in case of conversion from voice input characters to Kanji. CONSTITUTION:A voice input from a microphone 1 is recognized by a voice recognizing device 2, and the first, and the second, and the third candidates are outputted to a processor 3 in the order of distance. The recognized voice outputted to the processor 3 is converted to a hexadecimal as a word code in a Kanji dictionary by a code conversion table 3-1 provided in the processor. The word code of the first candidate is used as a key to retrieve a Kanji dictionary memory 3-2, and its contents are displayed on a display device 4 if it is found in the memory 3-2. If it is not found in the memory 3-2, word codes of the second candidate and the third candidate are used as keys to retrieve the memory 3-2 in order, and the input voice is converted to Kana (Japanese syllabary) and Kanji and is displayed on the device 4. Thus, the table 3-1 and the memory 3-2 are provided to automatize the retrieval operation in a certain degree, and the input speed is improved.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to an audio word processor that inputs characters by voice, proofreads the sentences created by the input characters, and performs all operations such as 1 mm group printing.

(Prior Art) Conventionally, this type of voice word processor has been used to input a word sound based on the operator's utterance, and then combine it with the operator's voice registered in advance in the voice recognition device (hereinafter referred to as standard voice) and human input. The degree of inconsistency with the given word sound is quantified as a digital value (hereinafter referred to as distance), and the first, second, and third candidates are output as recognition results in the order of the smallest distance. At this time, the recognized voice of the first candidate was displayed on the 1-table device, and the operator performed the kana-kanji conversion operation after confirming that it matched with his own input result.

(Problem to be solved by the invention) However, if the recognized voice of the input result does not match,
The operator's operations sequentially check the human results and each recognition result until they match, and until they match, the kana-kanji conversion cannot be executed, making the kana-kanji conversion operation complicated and slowing down the manual character processing.

Therefore, an object of the present invention is to enable an operator to search automatically to some extent for candidate characters in the voice recognition results when converting literary kana-kanji input by voice input, thereby increasing the input speed of a voice word processor. be.

(Means for Solving the Problem) According to the present invention, all the priorities are added to the recognized word code group generated as the recognition output of the voice input by one person in the order of matching or similar to the registered voice, and this code is When searching the entire dictionary as a search key, if the corresponding code is not in the dictionary,
A voice word processor is obtained which is characterized in that search keys are extracted one by one from the code group according to the priority and the search is performed sequentially.

(Example) Next, the present invention will be described in detail with reference to the drawings showing an example of the present invention.

In FIG. 1, the operator says the word he wants to input through the microphone l. The input word voice is sent to the voice recognition device 2.
After being encoded, the data is output to the processing device 3 and its contents are displayed on the display device 4. The voice input by voice input can be specified as kana, katakana, alphanumeric characters, symbols, etc. by selecting the mode of a group of input auxiliary keys on the keyboard 5.

Normally it is in Kana mode. When converting into kanji, pressing the conversion key on the keyboard 5 converts the content displayed on the display device 4 into kana-kanji. The sentences at the final stage after manual editing and editing are sequentially recorded in a single document record file 6. Further, when printing all of the recorded text, if all print keys on the keyboard 5 are pressed, the text will be printed on the printer 7.

FIG. 2 shows the code conversion table 3-
This is a diagrammatic representation of the outline of 1. The recognized speech output from the speech recognition device 2, ie.

An address whose reading of the kanji word in FIG. 2 (for example, Ao) matches the first, second, and third candidates is accessed.

segment number and word number) is 1 as the word code
It is converted into hexadecimal and sent to the kanji dictionary memory 3-2.

FIG. 3 shows the kanji dictionary memory 3-2 (
For example, the diagram illustrates the configuration of a RAM (RAM or ROM) (hereinafter referred to as dictionary memory). That is,
The first, second, and third candidates output from the speech recognition device 2 are converted into word codes of the kanji dictionary in the code conversion table 3-1 of FIG. Word code characters are read.

Furthermore, referring to FIG. 4, the operation of one embodiment of the present invention will be described. First, we will explain the case where all new voice words are registered in the voice recognition device 2 (step 42).
.

When the kanji to be registered is displayed on the display device 4 (step 54), the operator inputs the spoken word to be registered through the microphone 1 (step 55).

The input voice is analyzed into voice components by the voice recognition device 2 and converted into digital values (Step 561) The converted voice is stored in the data buffer 2-1 of the voice recognition device 2 as standard voice. (Step 57).

Next, the operation when the recognized speech is converted into Kanji will be explained. The operator first puts the device into input mode by operating the keyboard (step 42). When the operator speaks the word he wants to input into the microphone 1 while looking at the four display devices (step 43), the nine uttered words are displayed on the display device 4. When the operator presses the conversion key, the input voice is recognized by the voice recognition device 2 and output to the processing device 3 in the order of the first candidate, second candidate, and third candidate from the one with the smallest distance ( Step, Pua → Step 4
5→Step 46). The recognized speech outputted to the processing device 3 is converted into hexadecimal numbers as word codes of the kanji dictionary by the code conversion table 3-1 shown in FIG. 2 (step 47). Next, the dictionary 1 memory 3-1 in FIG. 3 is searched using the first candidate word code as a key, and if the word code exists in the dictionary, the contents are displayed on the display device 4 (step 49→step 50). →Step 52). If it is not in the dictionary, the dictionary memory 3-1 is searched using the word code as a key in the order of the first candidate and then the second candidate (step 50 →
Step 51 → Step 48 → Step 49). Once the input speech has been converted by kana-kanji conversion and displayed on the display device 4, the operator checks all the displayed contents, and if it is the desired word, the input is finished (steps 53). Also, the first candidate.

If there is no word in the kanji dictionary that corresponds to either the second or third candidate, the entire line is input again.

Steps 43 to 53 are executed.

As an example, a case will be explained in which the word "Aoi" (Aoi) is converted into kana-kanji.

When the character "ao" has already been input using the previous voice input, next say "i" and select " as the first candidate.
``Hi'', ``Good'' as the second candidate.

Furthermore, it is assumed that a recognition result of "a" is obtained as the third candidate. The processing device 3 accesses the code conversion table 3-1 in FIG. 2, and converts "Ao" and the first candidate "Hi".
The combination ``Aohi'' is searched. Since "Aohi" does not exist in the conversion table of FIG. 2, the combination "Aoii" of "Aohi" and the second candidate "ii" is searched.

This also does not exist, so finally the combination "Aoi" of "Ao" and the third candidate "I" is searched. "Aoi"
is in the code conversion table 3-1, so the processing device 3 displays “Mallow” on the display device 4. Here, when the operator fully presses the conversion key for kana-kanji conversion, the processing device 3 generates the conversion table shown in FIG. 2 to obtain the segment number and word number corresponding to "Aoi". The upper 4 bits of the address are generated from the obtained segment number, and the lower 8 bits of the address are generated from the word number. At this time, the segment number remains 1.
The word number is converted to hexadecimal to form the upper 4 bits of the address, while the word number is multiplied by 4 and converted to hexadecimal to form the lower 8 bits of the address. In other words, the address of "Aoi" is segment number 05. Since the word number is 02, it is 010100000010 in binary
, and is converted to 0508 in hexadecimal (however, the first 0 is a dummy number). Address 0508 of the dictionary memory 3-2 in FIG. 3, which corresponds to the converted word code, is accessed, and "Aoi" is read out as the kana-kanji conversion result.

The operator confirms everything and that it has been converted correctly.

Press all the no-conversion keys to complete the input. Furthermore, if a homophonetic kanji is required, the conversion key is pressed again to perform kana-kanji conversion.

Note that in this embodiment, the third candidate matches "Mallow" in the conversion table, but if the operator wants to display the first or second candidate, the operator can press the correction key fully and the next candidate will be displayed. Configure it so that

(Effects of the Invention) As described above, the present invention eliminates the need for the operator to search for recognized speech and recognition candidates through key operations when converting voice input into kana/kanji.
It simplifies the operation of inputting kanji characters and has the effect of completely increasing the speed of one-person input.

[Brief explanation of drawings]

Fig. 1 is a professional diagram of an embodiment of the present invention, Fig. 2 is a schematic diagram of a code conversion table, Fig. 3 is a configuration diagram of a kanji dictionary memory, and Fig. 4 is an operation of an embodiment of the present invention. It is a flowchart which shows. 1...Microphone, 2...Voice recognition device, 2
-1... Data buffer, 3... Processing device, 3-1... Code conversion table, 3-2
... Kanji dictionary memory, 4 ... Display device,
5...Keyboard, 6...Document record file, 7...Printer, 41-58...
70 - Chart steps. Agent Patent Attorney Susumu Uchihara ('” ” ”. Power Official Residence -1 No. 01 Q2 03 04 -
--Figure 2

Claims (1)

[Claims] A priority is added to the recognition word code group generated as the recognition output of the input speech in the order of matching or similar to the registered speech, and when searching a dictionary using the code as a search key, the code is used according to the priority. A spoken word processor characterized in that one of the groups is selected as a search key and the dictionary is sequentially searched.