JPS5852738A - Converting method of voice input japanese type writer - Google Patents

Abstract

PURPOSE:To input data with a pronunciation which is as natural as possible without increasing the number of recognization voices of a voice recognition part, by voicing a character string while prolonged symbols of the string are substituted by characters of vowel parts of character right before the prolonged- sound symbols. CONSTITUTION:In an independent word dictionary, prolonged-sound symbols in a KANA (Japanese syllabary) index field are converted into vowel characters of character right before them. Namely, prolonged-sound symbols after ''chi'' are all converted into ''i'', those after ''no'' are converted into ''o'', and those after ''ar'' are converted into ''a'', but an array of characters is not changed at all. Consequently, an automatic KANA-to-KANJI (Chinese character) conversion processing part performs conversion by following the same procedure with conventional KANA-to-KANJI conversion processing by key input, etc.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a voice input Japanese typewriter using a voice recognition device.

With conventional Japanese typewriters, when creating general documents, characters such as kanji, kana characters, alphanumeric characters, symbols, etc. are all arranged in parallel, and an operator selects characters from a table and presses the keys. . With this method, there were several thousand types of characters on the table, so Japanese typewriters were not devices that anyone could use immediately like English typewriters, and required the training of specialized operators.

In response, a Japanese typewriter has recently been proposed that uses a keyboard with only kana, alphanumeric, and symbols to input kana characters according to the pronunciation of the characters in the input sentence, and has the function of converting kana to kanji. .

Various kana-kanji conversion methods have been proposed, ranging from a simple dictionary lookup machine that displays kanji that correspond to the kana in a range specified by the operator, to automatic conversion of sentences by inputting the text as is. There are also tools that perform grammatical and semantic analysis and automatically convert kana to kanji. In the case of this automatic kana-kanji conversion device, the device includes a word dictionary and a grammar dictionary corresponding to the Japanese language dictionary.

A Japanese typewriter equipped with this automatic kana-kanji conversion function requires much less learning to input than conventional Japanese typewriters, and anyone can input.

However, in this case as well, since a single document is input, K becomes 1 if Kana characters, Kana lowercase characters, alphanumeric characters, symbols, etc. are input.
Since 40 to 150 input keys are required, a voice input Japanese typewriter using a voice recognition device in the input section is considered as a device that can obtain a certain degree of input speed to increase the input speed.

That is, a voice recognition device is used in place of the kana keyboard of the Japanese typewriter having the automatic kana-kanji conversion function. ``Various types of speech recognition devices have been proposed. However, when recognizing speech, it is divided into word recognition and monosyllable recognition, depending on the target. Among these, when considering the application of monosyllable recognition to voice typewriters, in principle it is possible to recognize an infinite number of languages because the target is a monosyllable.

A monosyllable recognition device is described in the Acoustical Society of Japan's "Monosyllable Speech Typewriter (Hokuten Applied Electrical Research Institute)" and is schematically shown in FIG.

In Figure 1, 1 is a microphone, 2 is a preamplifier, and 3 is a 16-channel bandpass filter (hereinafter referred to as 16C).
4 is a 16-channel analog-to-digital converter (hereinafter referred to as 16 CII A/D converter), 5 is a read-only memory for logarithmically converting input values (hereinafter referred to as logarithmic conversion ROM), 6 is a pitch extractor, 7 is a counter timer, and 8 is a microprocessor C.
(hereinafter referred to as micro cPU), 9 is the pass line (hereinafter referred to as B
10 is a read-on memory (hereinafter referred to as ROM), 11 is a random access memory (hereinafter referred to as RAM), 12 is a floppy disk drive device (hereinafter referred to as FDD), 13 is a character display device (hereinafter referred to as CRT), 14 is a keyboard device (hereinafter referred to as KB), and 15 is a hardware arithmetic unit.

The voice typewriter having the above configuration operates as follows. First, since the audio input from microphone 1 is attenuated by 6 dB/octave, preamplifier 2 emphasizes the high range at a rate of 6 dB/octave, flattening the power spectrum and increasing its output to 16 dB/octave.
Input to CHBP3, divide the frequency from 200 to 4400Hz into 16 channels, and send the output to A/D for each CH.
Enter Konnoku Ichiyo 4. 1 at A/D converter 4
After converting to 2-bit digital quantity, logarithmic conversion R, OM
5 and convert to 8-bit pairs. This is to approximate the logarithmic nature of human auditory characteristics, and
Furthermore, this process takes into consideration the advantage that when normalizing the power spectrum, calculations can be performed using only addition and subtraction. The output of the A/D converter 4 is drawn into the micro CPU 8 through the BUS LINE 9 of the micro CPU 8. The micro CPU 8 includes a ROM 10 for control, a RAM 11 for a data area, an FDD 12 for storing data such as registered voices, and a voice recognition result output display (TITI).
3. A KB 14 for manually inputting information to the micro CPU 8 is connected to constitute a small electronic computer system.

Further, the output of the 16 CH BPF 3 is input to the pinch extractor 6, and the configuration is configured to trigger the 16 CHA/D converter every pinch period. One method for the pitch extractor 6 is proposed in, for example, Japanese Patent Application Laid-Open No. 162405/1988 entitled "Pitch Frequency Extraction Apparatus", and it may be considered that the pitch frequency is extracted by such a method. A pitch frequency exists in the case of a vowel or a voiced consonant, but in the case of a voiceless consonant, the sound source is noise and the pitch frequency does not exist. In order to sample these unvoiced consonants, the counter timer 7 is used at a fixed period, for example, l Kl (1 m se at the frequency of z).
A pulse is generated every c and inputted to the 16c HA/D converter 4 -f.

The 16CHA/D converter 4 samples the output of the 16CH BPF 3 every pitch period in the case of a vowel or voiced consonant, and every fixed period in the case of a voiceless consonant, and performs analog-to-digital conversion to obtain the power spectrum of the input speech waveform. Enveloping properties can be obtained.

The digitized human voice waveform obtained in the above manner is subjected to power spectrum normalization, and then the squared distance W is calculated between it and each data of the registered voice waveform.

Calculation of this squared distance requires several thousand times, and the micro CP
Since real-time processing cannot be performed by calculating with U8,
The hardware operator 15 performs calculations on a hardware basis.
The recognized monosyllables are sequentially outputted as a single word or phrase and input into the automatic kana-kanji conversion device at one point, where they are converted into a sentence containing kanji and kana.

For this kana-kanji conversion, please refer to "Research Practical Report Vol. 26, No. 1 (January 1977) published by Nippon Telegraph and Telephone Public Corporation Telecommunications Research Institute, p353-369 Prototype of online kana-kanji conversion system (Kimura et al.)" A known method such as shown in can be used.

FIG. 2 shows an example of the automatic kana-kanji conversion processing procedure.

Briefly explaining the processing procedure shown in FIG. 2, if a kana sentence separated into phrases is input, human phrases are processed by the longest match method. The longest match method is a method in which an independent word with the longest word length is detected in a bunsetsu unit by comparison with a dictionary headword, and the input bunsetsu is separated into an independent part and an attached part. Through this process, if the separated adjunct can be further separated, it is separated, and then a grammar matching process is performed to test whether connection with the separated independent word or connection between adjuncts is possible, and if separation is appropriate. If it is inappropriate, it moves to the next process, and if it is inappropriate, it returns to the longest match process again, detects the next independent word with the longest word length, and performs the same process. If the grammar matching process determines that it is appropriate, the separation of the input phrase into independent words and dependent words is completed.

Next, if the separated independent words have intersyllabic allographs, the semantic information of the homophones that are candidates for separation and selection is extracted from the dictionary, and the semantic information of the homophones is extracted from the dictionary to find the most semantically Performs semantic processing to select items that are highly connected. In addition, if the candidate homophones cannot be narrowed down to one by this process alone, 1.11 extract the frequency information of the candidate homonyms and select the one with the highest frequency as the final one. It performs output processing of clauses as an independent #ri.

In the above processing, longest match processing and grammar matching processing are performed (・If all of them are non-conforming, there is a possibility that a prefix exists at the beginning of the clause, so post-prefix processing is performed to check the prefix of the clause. All you have to do is remove the part that seems to be the same and perform the longest match process again.

Here, we take "Onseinori" and "Team Color Karamo" separately as specific inputs, and the processing procedures such as longest match processing, grammar matching processing, prefix processing, etc. and the processing results are shown in Tables 1 and 2, respectively. .

Table 1"""°゛'';8a'l l! <':'GJ'
2 As shown in the table above, by connecting a monosyllabic speech recognition device and an automatic kana-kanji conversion device, a voice input Japanese typewriter that converts words or phrases input by voice into sentences containing kanji and kana and outputs them. Although it is possible in principle to make
It has serious drawbacks as follows.

That is, the kana characters in Japanese are phonetic characters, each of which has one independent +-t,'- utterance. Therefore, generally speaking, the sentence or words to be input can be converted into kana characters and then uttered literally. in this case,
The symbol used is the long symbol "-". This symbol itself does not have an independent phonation, but only serves to extend the utterance of the character immediately before it.

In the case of conventional automatic kana-kanji conversion devices, this was handled by inputting the "-" symbol from the keyboard, but when considering countermeasures for the voice input method, the first thing to do is to input the symbol "-" from the keyboard. It may be possible to improve the monosyllable recognizer so that it can distinguish between sounds without an "-" symbol, but in that case, "a"
and "Ah", "I" and "E", etc. Since it is necessary to recognize twice as many characters as below, the amount of reference data increases, and the memory capacity to store it also increases. This is not a good idea because it increases the cost, increases the recognition processing time, and increases the number of erroneous recognitions due to the similarity of the generated sounds. Another possibility is to assign a special generated sound to the "-" symbol.

For example, if you create and assign a single continuous sound called ``Chou'', it can be treated in the same way as input from a keyboard.

However, in such a case, the inputs such as "control" or "team" are
Because it is necessary to pronounce words such as "ru" or "chi, chou, mu," etc., the sense of incongruity between the character string and the utterance increases, making it difficult to utter and diminishing the features of voice input.
It's a good idea. Another method is to omit the utterance of the "-" symbol and enter the input. The above-mentioned ``control'' ya 1 team'' becomes ``control'' ya ``chim.'' In this case, although it is not difficult to utter, since some of the input information is omitted, there is naturally a risk that errors will increase due to the reduction in input information. In other words, "car" is 1
``ka'', ``nurse'' become ``eggplant'', and ``ace'' become ``nis'', making it easy to distinguish them from other words.
−” This cannot be said to be a solution to the problem with the symbol.

The present invention is to compensate for the shortcomings of these improvement measures, without increasing the number of sounds recognized by the speech recognition unit (and by making it possible to input a given character string with a pronunciation that is as close to natural as possible). This method provides a kana-kanji conversion method for voice input.In other words, when inputting voice, if there is a long sound symbol "-" in the character string, "-" is used instead of this "-".
The vowel part of the character immediately before the symbol will be uttered, and for every "-" symbol that appears in the headword of the independent word dictionary used for automatic kana-kanji conversion processing, the character immediately before that symbol will be uttered. It is converted into the vowel part of the character,
Automatic kana-kanji conversion processing is performed using an independent word dictionary that leaves other matters, such as notation and parts of speech, as they were. Therefore, the conversion is the same for both utterances and headwords. For example, ``contronil'' becomes ``control perle,'' and ``chinim'' becomes ``chimimu.''

This converted utterance is different from long-sound utterance, but in monosyllabic utterance there is almost no sense of discomfort and it is close to natural.

Also, the order in which the headwords are arranged by converting the dictionary headwords varies depending on the dictionary used, but for example,
In the case of Gakken Japanese Dictionary by Gakushu Kenkyusha, the processing for the headword arrangement of the "-" symbol is as follows: [Long sounds are placed in the position where either A, I, Tsu, Engineering, or O is applied according to the pronunciation ( .Example burner → Baanaa''. In a dictionary with such an arrangement, the arrangement remains unchanged in the character conversion according to the present invention.

An embodiment of the present invention will be described below.

Figure 3 shows an example of a general independent word dictionary.
FIG. 4 is a diagram showing an example of an independent word dictionary used in the present invention.

Dictionaries used for automatic kana-kanji conversion have some kind of coded data stored in various storage devices, but the dictionary entry characters are JIS-1 due to their capacity and simplicity.
6220 code is used, and the JIS-6226 code, which can represent Kanji characters, is generally used as the notation code.

Note that the codes and formats of other information such as part of speech and frequency information differ depending on each system. In the independent word dictionary created in this way shown in FIG. 3, the long sound symbol r-J on the kana heading is converted into the vowel character of the character immediately before it, and the independent word dictionary shown in FIG. 4 is obtained. In the example in this list, the relevant heading K is marked with a seal, but all the "-" symbols after "chi" are converted to "i", and the symbols after "]" are converted to "i". The "-" symbol is converted to "o", and the "-" symbol after "u" is converted to "a". However, the word arrangement has not changed at all. Furthermore, when the present invention is implemented, all "-" symbols on headings in the affix dictionary are also converted. For example, "all Japan" or "all transistor"
It is convenient to treat ``all'' as a prefix when used in expressions such as ``all''.

If this ``oruko'' is included in the affix dictionary, the heading will be converted to ``oru''. Note that in this case as well, there is a notation section in addition to the kana headings, as in the independent word dictionary, and the notation section is in JIS-6226 code. Also, 1- on the notation section
The symbol shall be preserved as is.

Table 3 shows the processing procedure and processing results when the character string "from team color" is input using the dictionary according to the present invention.

Table 3 In this case, the voice input for [from team color] is ``Team Kara Akaramo'', which can be uttered without feeling too strange to the original character string. Next, a verification process is performed using this as input. However, even if the dictionary heading is converted according to the pronunciation, the notation is the original "-"
” symbol is preserved, so if there are no errors in verification and decomposition, the final output will be the desired “from team color”.

As is clear from the above explanation, according to the present invention, the "-" symbol, which does not have its own pronunciation, is replaced with the vowel character of the character immediately before the symbol. Even if you make a voice input that doesn't feel too strange, the automatic kana-kanji conversion processing section will stop at the conventional kana-kanji conversion process using manual input (it is possible to process using the same processing procedure).

Here, when the present invention is implemented, some headings that were conventionally distinct from each other may become the same on the headings. For example, for the heading ``Chiiku'' in FIG. 4, the notations ``Education'' and ``Cheek'' exist. The appearance of many so-called homonyms is an undesirable phenomenon in order to obtain correct notation, but for example, if we take Shogakukan's Newly Selected Japanese Dictionary as an example, the total number of heading characters is about 7.
Compared to 10,000 homonyms, the number of homonyms due to the heading conversion of the present invention is only increased by about an addition point, and among them, fugue...
...[Interpretation] Fugue (homophone) Teutonic... [Interpretation] One of the Germanic tribes (homophone) You can omit either word in special words such as garrison. This level of increase in homonyms is not a problem in practical terms.

In addition, when inputting something other than voice input into the automatic kana-kanji conversion processing unit having a dictionary group implementing the present invention, it is also possible to input a "-" symbol, but in that case, the input "-" symbol will be replaced with the "-" symbol. If the symbol is uniquely converted to the vowel character of the character immediately before it, processing will occur without any problem. In addition, if converting the "-" symbol is complicated, leave the previous heading section containing the "-" symbol and create a new section with the "-" symbol converted to a vowel character. You can also think of ways to add it.

However, since this method involves adding items, it is necessary to increase the storage capacity of the memory.

In any case, in the automatic kana-kanji conversion device implementing the present invention, it is possible to perform the conversion process for the input of the "-" symbol as described above, or by saving the entry of the "-" symbol, compared to the conventional keyboard. It is also possible to process both raw character string input and voice input at the same time.

[Brief explanation of drawings]

FIG. 3 is an example of a general independent word dictionary, and FIG. 4 is a diagram showing an example of an independent word dictionary used in the present invention. 1...Microphone, 2...Preamplifier, 3...+6CHBPF.

Claims (1)

[Claims] In the kana-kanji conversion method that converts and outputs a kana character string input by voice into a sentence containing kanji and kana, an electronic system that replaces the long sound symbol r-J that appears on the kana heading with the vowel character of the character immediately before this symbol. It is characterized by comprising an input character verification dictionary configured by a device, comparing the kana heading of the input character verification dictionary with a kana character string input by voice, and outputting a notation corresponding to the input kana character string. A method for converting kana to kanji in a voice-input Japanese typewriter.