JP2002123507A - Device and method for pronouncing chinese and converting chinese character - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain KANJI (Chinese character) candidates including desired KANJI even when no tune or a wrong tune is inputted. SOLUTION: Input data in Pin Yin pronunciation and input data in Zhu Yin pronunciation are converted into corresponding sound codes by using a Pin Yin/sound code conversion table 31 and a Zhu Yin/sound code conversion table 32 respectively. A dictionary 35 contains KANJI codes (corresponding to words) so that they correspond to sound code strings. A sound code string is generated from input data. The respective sound codes of the input sound code string and the sound codes of a sound code string in the dictionary 35 are compared with each other through a filter which masks specific bits of sound codes. Through this comparison, the KANJI code corresponding to the matching sound code string is read out of the dictionary 35 and the word (KANJI) corresponding to the KANJI code is displayed.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus and a method for converting Chinese phonetic notation input from a keyboard into corresponding Chinese characters and outputting the converted Chinese kanji, and is suitably used particularly in a Chinese word processor or work station. Apparatus and method.

[0002]

2. Description of the Related Art Chinese is expressed using Chinese characters. There are several notations for writing the pronunciation of kanji. A typical example was the Chinese government's promulgation in 1958.

(Equation 1) Used before and still in Taiwan

(Equation 2) There is.

[0003] Each Chinese kanji character is pronounced by a vowel (Sheng Mu) corresponding to a consonant, a rhyme (Yun Mu) corresponding to a vowel,
Four tones (Si Sheng) or tone (Sheng Dia)
o) and can be divided into Shy Yun is a combination of a mother and a mother. Some kanji pronunciations have no intonation. The pronunciation of one kanji is represented by one or less (one or zero) initials and one rhyme (and further using the tone, if necessary).

There are four types of tones. One voice (Yi Sheng or 1 Sheng): Treble and flat.

(Equation 3) Two voices (Er Sheng or 2Sheng): rising from bass to treble.

(Equation 4) Three voices (Shan Sheng or 3Sheng): From high notes to low notes, again to high notes.

(Equation 5) Four voices (Si Sheng or 4 Sheng): descending from treble to bass.

(Equation 6)

For example, the Chinese character "China" is Pin Yin
According to the law,

(Equation 7) Is written. Among these, "Zh" and "G" are the initials, and "ong" and "uo" are the rhymes. Also,
According to the Pin Yin method, the kanji "Japan"

(Equation 8) Is written. “R” and “B” are initials, and “i” and “en” are rhymes.

In a conventional Chinese word processor, Pin
Only input in Yin notation was allowed. Because the Pin Yin method is relatively new, we know the Zhu Yin method but Pin Yi
n There are people or generations who do not know the law. Therefore, if you want more people to take advantage of the Chinese word processor, you must also allow Zhu Yin input.

[0007] In addition, the Pin Yin method notation uses Mandarin as a standard language. In the vast China, there are languages that have different tones than Mandarin. Depending on the region, even the voice humor may be different from Mandarin. Therefore, it is difficult for people who do not know or are familiar with Mandarin as the standard language to input vocal rhythms and tones correctly.
Input errors often occur. Even in Mandarin-speaking people,
The pronunciation is not always conscious of tone distinction, and the input operation to the word processor must be performed while remembering or thinking about the tone. There are times when you can not enter.

In the conventional Chinese word processor, the correct kanji corresponding to the vocal rhythm and tone is output only when the syllable and tone are correctly input, and a correct kanji cannot be obtained if there is an input error.

[0009]

DISCLOSURE OF THE INVENTION An object of the present invention is to provide a candidate kanji including a desired kanji even if the tone is not input or the tone is incorrectly input.

Another object of the present invention is to provide a kanji candidate corresponding to a pronunciation including a partial pronunciation if at least a part of the pronunciation is correct.

A Chinese phonetic notation / kanji conversion device according to the present invention is a conversion means for converting data representing an input Chinese pronunciation into a sound code corresponding to the pronunciation, a sound code, and a sound code. A dictionary in which kanji codes indicating kanji having pronounced pronunciations are stored in association with each other,
Filtering means for masking one or more predetermined bits constituting the sound code, by comparing the sound code obtained from the conversion means with the sound code of the dictionary using the filtering means and comparing each other with each other, There is provided control means for searching the dictionary for a sound code that matches the sound code obtained from the conversion means, and reading a kanji code corresponding to the matching sound code from the dictionary.

[0012] In one embodiment, the sound code is
It is configured to include a bit indicating a voice, a bit indicating a rhyme, and a bit indicating a tone. In this case, the filtering means is configured to mask a bit representing the initial, a bit representing the pronoun or a bit representing the tone.

It is to be understood that the filtering means includes a means for passing a sound code as it is.

A search mode selecting means for selecting whether or not the filtering means is used or selecting one of the plurality of filtering means is further provided as necessary.

[0015] In one embodiment, the conversion means is a speech recognition means for recognizing pronunciation based on a speech input signal and outputting a sound code corresponding to the pronunciation.

Means for converting a read kanji code into display data representing a kanji represented by the kanji code to realize a more realistic word processor, and a device for displaying kanji based on the display data There will be provided a designation input means for designating any one of the displayed kanji candidates, and a memory for storing a kanji code indicating the designated kanji.

Further, while the conversion means is configured to convert input data for one kanji into a sound code, the dictionary stores a sound code string and a kanji code for a word consisting of one kanji or a plurality of kanji. Are stored in association with each other. Then, a series of input data is converted into sound codes by dividing each input data into one kanji, and one or a plurality of converted sound codes are arranged in word units to form a sound code string. A kanji code corresponding to this sound code string is searched in the dictionary.

According to the present invention, the sound code representing the input data and the sound code in the dictionary are both filtered and then compared. Since a portion (one or a plurality of bits) of a sound code to which a match or a mismatch should be given is masked by the filter, the masked portion is not subjected to comparison processing.

Therefore, when it is desired to make the tone unquestionable, a filter suitable for the tone can be used, regardless of whether the tone matches or does not match, or without inputting the tone.
One or more kanji candidates corresponding to the sound code corresponding to the sound code corresponding to the input data in the other part (voice rhythm) are obtained. In this way, even if the tone is not input or the tone is incorrectly input, the candidate of the kanji (word) including the desired kanji (word) is output.

The type of filter can be set arbitrarily.
Therefore, it is also possible to search for kanji under the condition that only the initials match or only the rhymes match. That is, if at least a part of the pronunciation is correct, a candidate for a kanji (word) corresponding to the pronunciation including the pronunciation can be obtained.

The present invention further provides a Chinese phonetic notation / kanji conversion method corresponding to the above-described Chinese phonetic notation / kanji conversion device.

In the Chinese phonetic notation / kanji conversion method, a dictionary is prepared in which sound codes and kanji codes indicating kanji having a pronunciation represented by the sound codes are stored in advance in correspondence with each other, and input is performed. The data representing the pronounced Chinese pronunciation into a sound code corresponding to the pronunciation,
The sound code obtained by the conversion and the sound code of the dictionary are filtered by masking one or more predetermined bits constituting the sound code, and are compared with each other. A matching sound code is searched in the dictionary, and a kanji code corresponding to the matching sound code is read from the dictionary.

Other features and advantages of the present invention will become apparent in the following description of embodiments with reference to the drawings.

[0024]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows the configuration of a Chinese phonetic transcription / kanji conversion device. The apparatus will generally be implemented as part of a Chinese word processor, work station, or the like.

The Chinese phonetic transcription / kanji conversion device includes a computer 10 including a central processing unit (CPU), a phonetic transcription,
Keyboard for entering various modes and other functions
20, a memory device that stores a dictionary and various conversion tables
30 and a display device 14 for displaying converted kanji characters, other information or data, and a control device 12 thereof.

As the computer 10, a commercially available general-purpose computer can be used. The computer 10 is programmed to execute input, edit processing and kanji search processing, which will be described later.

In this embodiment, there are a Pin Yin method and a Zhu Yin method as phonetic notations that can be input. In addition, a complete phonetic notation including the tone and a phonetic notation performed only in the vocal rhythm excluding the tone are allowed.

The keyboard 20 has a Pin Yin key 21 for inputting phonetic notation by the Pin Yin method, a Zhu Yin key 22 for inputting phonetic notation by the Zhu Yin method, and a phonetic notation to be input by the Pin Yin method. Input mode key 23 for selecting whether to use the Khu or Zhu Yin method. I want you to search the kanji using the complete phonetic notation including the tone (this is called the first mode), or use only the voice rhythm without the tone. A conversion mode key for selecting whether to search for kanji using phonetic notation (this is referred to as a second mode), a conversion key for instructing that the entered phonetic notation be converted to kanji,
A space key (if necessary) and a function key 25 for inputting other functions are provided.

The data (or chord) representing the phonetic spelling input from the keyboard 20 corresponds to the corresponding sound code (Yin
Code). Pin Yin for this transcoding
A / sound code conversion table 31 and a Zhu Yin / sound code conversion table 32 are provided in the memory device 30. To display the entered phonetic transcription in its phonetic transcription or another phonetic transcription, once converted to a phonetic code, use Pin Yi
Sound code / Pin Yin table 33 for reverse conversion to data (or chord) representing n notation or Zhu Yin notation
And a sound code / Zhu Yin table 34 are provided in the memory device 30. Further, a dictionary 35 used for searching for a code (kanji code) representing a corresponding kanji based on the converted sound code, and a converted data area 36 for storing the converted kanji code are provided in a memory device. 30 are provided. The memory device 30 is realized by a semiconductor memory (ROM or RAM), a magnetic memory (floppy disk or hard disk), or a combination thereof. For example, the conversion tables 31 to 34 are stored in a ROM, a floppy disk or a hard disk, the dictionary 35 is stored in a floppy disk or a hard disk, and the converted data area is set in the RAM.

As the display device 14, most commonly, C
Although an RT display device is used, a plasma display device or a liquid crystal display device can also be used. The display control device 12 has a built-in character generator 13. Character generator 13 is Pin Yin or Zh
u Converts data representing Yin and kanji codes to display data (dot data).

FIG. 2 shows the arrangement shown in FIG. 1 arranged according to its function and processing flow. Switching circuit 1
5. The editing processing device 16 and the kanji search processing device 17 are substantially realized by the computer 10. Alternatively, the phonetic transcription / kanji conversion device may be configured to have the hardware architecture shown in FIG.

The switching circuit 15 is provided with a Pin Yin key 21 or a Zhu Yin key 22 according to a selection input through an input mode key 23.
Select Data representing the Pin Yin or Zhu Yin notation input using the Pin Yin key 21 or the Zhu Yin key 22 is provided to the editing processing device 16. Switching circuit 15 is Pin
It is also possible to automatically discriminate between a Yin key input and a Zhu Yin key input and select the input.

Some Chinese words are composed of one kanji, and some are composed of a plurality of (typically two or three) kanji. The editing processing unit 16 separates the input data string representing the Pin Yin or Zhu Yin notation into data representing one kanji character, and converts the separated data into the conversion table 31 or 32 selected by the input mode key 23. Refer to and convert to sound code. When there is an input from the conversion key included in the function key 25, the sound code string created from the data input up to that time is provided from the editing processing device 16 to the kanji search processing device 17. Since the edit processing device 16 supplies the data representing the input Pin Yin or Zhu Yin notation to the display control device 12, the input Pin Yin or Zhu Yin notation character is input on the screen of the display device 14. They are displayed in order. In this case, the reverse conversion tables 33 and 34 are not required. After the edit processing device 16 converts the input data into a sound code, it is changed to Pin Yin or Zhu Yin.
When performing notation display, the inverse conversion tables 33 and 34 are used. These reverse conversion tables 33 and 34 are, in particular,
Data input by the Pin Yin method (or Zhu Yin method)
It is used effectively when displaying in Zhu Yin notation (or Pin Yin notation). This is convenient because an operator who only knows Pin Yin notation can know Zhu Yin notation, and an operator who knows only Zhu Yin notation can know Pin Yin notation.

The kanji search processing device 17 searches the dictionary 35 in accordance with the selection by the conversion mode key 24 based on the given sound code string, and obtains one or more kanji having a pronunciation represented by the sound code string. Is read out and given to the display control device 12. The display control device 12 reads display data for displaying the kanji represented by the given kanji code from the character generator 13 and, based on the read data, displays one or more kanji (kanji candidates) on the display screen of the display device 14. Is displayed. When the operator checks the displayed kanji using the function keys 25 or selects any one of them, the kanji code representing the kanji confirmed or selected is stored in the converted data area 36.

FIG. 3 shows an example of the Pin Yin / sound code conversion table, and FIG. 4 shows an example of the Zhu Yin / sound code conversion table.

For one type of pronunciation, the Pin Yin method and Zhu Yi
There are various phonetic notations, such as the n method, but these notations always correspond to one type of pronunciation. One type of sound code (Yin Code) is assigned to one type of pronunciation. One Pin Yin notation always corresponds to one sound code, and one Zhu Yin notation always corresponds to one sound code. Pin Yin notation for one pronunciation and the same pronunciation
The Zhu Yin notation corresponds to one common chord. In this way, even if there are a plurality of types of pronunciation notations used for inputting pronunciations, the same pronunciation is expressed using one sound code. Regardless of the input using the Pin Yin method or the input using the Zhu Yin method, the same sound is converted to one sound code, so the sound code is the only sound code inside the device. Can be uniformly used as a code representing. For this reason, there is no need to prepare a dictionary for each phonetic notation, such as a dictionary for the Pin Yin method and a dictionary for the Zhu Yin method, and dictionaries that are searched using phonetic codes common to all notations are used. It is enough if you prepare.

For example, the pronunciation in Pin Yin notation shown in the first line of FIG. 3 and the pronunciation in ZhuYin notation in the first line of FIG. 4 are the same, and thus correspond to the same sound code 52f8 (hexadecimal notation). The pronunciation of the other lines is the same. In FIGS. 3 and 4, the leftmost column is represented by Pin Yin notation and Zhu Yin notation, respectively, for the sake of clarity, but these notations are represented by binary data in the conversion table. Needless to say.

FIG. 5A shows the data format of a sound code. In this embodiment, the sound code is composed of 2 bytes. The upper 1 byte mainly indicates the mother, and the lower 1
The bytes mainly represent the initial. The data “0” of the most significant bit (f) in the upper one byte and the data “1” of the most significant bit (7) in the lower one byte distinguish the upper one byte and the lower one byte constituting one sound code. At the same time, it is used to distinguish it from other data (especially other sound codes in a sound code string).

The least significant bit (eighth bit) of the upper one byte indicates the presence or absence of a tone. Some sounds have no tone. No tone is represented by "0", and tone presence is represented by "1". The lower 2 bits (0th to 1st bits) of the lower 1 byte represent a tone. As shown in FIG. 5 (B), one voice, two voices, three voices, and four voices are “00”, “01”, “10”,
It is represented by "11".

The middle 6 bits of the upper 1 byte (9th to e
(Bit) represents the prim, and the middle 5 bits (2nd to 6th bits) of the lower 1 byte represent the initial. Since there are 37 types of rhymes and 24 types of vowels, this number of bits is sufficient.

A filter is used in the kanji search processing described later. This filter also consists of 2 bytes,
The bit, the first bit and the eighth bit are set to “0”, and the other bits are set to “1”. If this filter is expressed in hexadecimal, it becomes "FEFC".

FIG. 6 shows the procedure of input and editing processing by the computer 10 or the operation of the editing processing device 16.
The computer 10 or the editing processor 16 is provided with a key data buffer as shown in FIG. 7 and a sound code string buffer as shown in FIG.

First, Pin Yi is pressed by the input mode key 23.
It is determined whether the mode of the n method or the Zhu Yin method is set (step 41). If the Pin Yin method is selected, the Pin Yin / tone code conversion table 31 is selected (step 42), and the Zhu Yin Zhu Yi if law is selected
n / sound code conversion table 32 is selected (step 4
3).

Subsequently, the first key is pressed by the conversion mode key 24.
It is determined whether the mode or the second mode is selected (step 44). When the first mode is selected, no processing is required, but when the second mode is selected, “FEFC” is set in the filter (realized by a register or a memory area). (Step 45). When the first mode is selected, data “FFF” consisting of all bits 1 is used as a filter.
F "may be set.

Each time character or symbol data representing a phonetic notation is input from the Pin Yin key 21 or Zhu Yin key 22, these data are stored in the key data buffer (step 47). As shown in FIG. 7, when one character is input, terminal symbol data “φ” is stored at the subsequent stage of the character. This is because the data representing phonetic notation is variable length data, and it is necessary to specify the end of the data. FIG. 7 shows a state in which the pronunciation of “Zhong” in Pin Yin notation is being input according to the second mode.

It is determined whether key data for one kanji has been input (step 48). There are various methods for this determination. The first is to make the operator press the space key at the end of key data input for one kanji. If there is an input using the space key, it is determined that the input for one kanji has been completed. The second is effective in the case of the first mode, in which the operator is required to input the tone using numbers 1, 2, 3, and 4 after inputting phonetic symbols. For example, a character having a pronunciation of “Zhong” and one voice inflection is input as “Zhong1”. If there is a numeric key input, it is determined that the input of one kanji data has been completed. The third is to make a determination by automatic recognition based on syllable division. Since there is a certain rule in the description of pronunciation by the Pin Yin method, by using this rule, it is possible to determine where in a key data sequence an expression of one kanji ends. Notation by the Zhu Yin method can be used because there are certain rules in the same way.

In any case, when key data for one kanji is input, the input key is referred to by referring to the previously selected PinYin / sound code conversion table 31 or Zhu Yin / sound code conversion table 32.・ Data (Pin Yin or Zh
u Yin notation) is converted to the corresponding sound code.
This sound code is stored in the sound code string buffer (step 49).

The above-described processing of step 47 is repeated until the input of key data for one kanji is completed (step 48), and the processing of steps 47 to 49 is repeatedly executed until the conversion key is pressed (step 46). ). Then, when there is a conversion key input, the sound code string stored in the sound code string buffer is passed to the kanji search process starting from FIG. 9 (step 50).

For example, when “Zhong Guo” is input in the second mode by the Pin Yin method, the key input data “Zhong” is converted to the sound code “52f8” and “G
“uo” is converted into a sound code “66b4”, and a sound code string “52f866b4” is obtained.

After designating the second mode, a pronunciation including a tone may be inputted. For example, "Zhong1 G
uo2 ". In this case, "53
A sound code string “f867b5” is generated. Since the second mode is specified, the filter “FEF”
"C" is set (step 45), and the second mode search described in detail later is executed.

FIGS. 9 to 11 show the procedure of the kanji search processing particularly in the second mode. This processing is also applied to the kanji search processing in the first mode as it is by setting the filter to “FFFF”. Also,
This process may be executed by the computer 10 shown in FIG. 1 or the kanji search processing device 17 shown in FIG.

Prior to the description of the kanji search processing, the dictionary 35
Will be described with reference to FIG. 12 and FIG. Figure
As shown in FIG. 12, the dictionary 35 is provided with an index I table, an index II table, and a sound code string / kanji code correspondence table.

As shown in FIG. 13, the sound code string / kanji code correspondence table stores a sound code string and a kanji code representing one or a plurality of kanjis constituting a word having a pronunciation represented by the sound code string. It is also stored. Although FIG. 13 shows the kanji itself instead of the kanji code for easy understanding, it should be understood that a code represented by a binary number is actually stored.

Since the word "Chinese" is composed of three Chinese characters, the sound code string is composed of 6 bytes. A word consisting of two kanji characters (for example, "China") corresponds to a 4-byte sound code string. One kanji corresponds to a two-byte sound code string. in this way,
Words that share the first kanji ("middle" in the above example) are arranged close to each other, and arranged so that the sound code string that constitutes the word has a larger number of bytes than the value indicating the relative address. Have been. In FIG. 13, a symbol "φ" representing "0000" always exists at the end of the sound code string.

One pronunciation may represent two or more Chinese characters. For example, the sound code strings of the relative addresses 102 and 103 are both "53f8", which correspond to the Chinese characters "chu", "chu", and the like.

The relative address of the sound code string / kanji code correspondence table is represented by p. The sound code string at the relative address p is represented by Y0 (p, 1), Y0 (p, 2),. Y0 (p, 1), Y0 (p, 2) and the like are generally represented by Y0 (p, C) (C = 1, 2,...). The kanji code of the relative address p is represented by KA (p) (variable length).

The sound code string / kanji code correspondence table stores as many words as possible (preferably, almost all words used in China). The order of arrangement of these words is free except for the rules described above. Therefore, a pair of an arbitrary sound code string and a kanji code can be arranged in an arbitrary storage location. Let M be the number of words arranged in the sound code string / kanji code correspondence table.

Referring back to FIG. 12, the index I table and the index II table store the sound code strings of the randomly arranged sound code string / kanji code correspondence table.
This is for enabling a search in the order of the numerical values represented by the sound code strings.

The index I table contains N In
dexI (i) is arranged in a certain order. Ind
exI (i) is a pointer to the corresponding element in the index II table (indicating the relative address of the index II table). N is the number of different sound code strings in the sound code string / kanji code correspondence table.
As described above, since two or more words can correspond to one sound code string, N ≦ M is generally satisfied.

The index II table has M storage locations, and each storage location has three types of elements, F1 (k) and F1 (k).
2 (k) and F3 (k) are stored. F3 (k)
Is a pointer to a corresponding sound code string in the sound code string / kanji code correspondence table (indicating a relative address of the correspondence table). F2 (k) is relative to another storage location in the index II table having F3 (k) pointing to the same sound code string as the sound code string pointed to by F3 (k) in the same storage location. Address). As a result, both the words “medium” and “chu” can be searched for the sound code string 53f8. If there is no other identical chord sequence, F2 (k) = φ is set.
F1 (k) is F3 (k) in the same storage location
F3 that points to a sound code string that includes the same sound code string as the sound code string to which F3 (k) points (ie, a sound code string that is longer than the sound code string to which F3 (k) points)
(K) Indicates another storage location (relative address) in the index II table. As a result, when "China" is searched, "Chinese" containing "China" and having more kanji characters will be automatically searched.

Inde in index I table
xI (i) is sorted in advance in ascending order of the numerical value represented by the sound code string in the sound code string / kanji code correspondence table. Therefore, even if the arrangement of the sound code strings in the sound code string / kanji code correspondence table is random, when viewed through the index I table, in the sound code string / kanji code correspondence table, the sound code strings represent It looks as if the numbers are arranged in ascending order.

In the kanji search processing shown in FIGS. 9 to 11, a binary search (binary search or dichotomizing) technique is used.

Several variables are used in the kanji search processing. These variables are START, END,
find and the like. The variables START and END are used to access the index I table.
find is used to point to the storage location of the kanji code buffer (see FIG. 14) that stores the found kanji code. These variables are implemented as data stored in registers or memory areas.

Input / edit processing (FIG. 6) or input sound code strings given from the edit processing device 16 to the kanji search process or the kanji search processing device 17 are represented by x (1), x (2), x
(3), ..., φ. For example, when “Zhong1 Guo2” is input in Pin Yin notation, the input sound code string is “53f8 67b5 φ”. That is, x (1) = 53f8, x (2) = 67b
5 A sound code counter C is used to indicate the position of each sound code constituting the input sound code sequence in the input sound code sequence. For example, x
(1) is represented by x (C) (C = 1).

In FIG. 9, first, a sound code counter C
Is initialized to 1 (C = 1, step 51). Thereby, the sound code x (1) at the head of the input sound code sequence is specified.

Subsequently, the variable START is initialized to 0, END is initialized to (N-1), and find is initialized to 0 (step 52).

Using the variables START and END, the relative address of the index I table is set to (START + EN).
D) / 2, which is set as i (step 54). This is a process for obtaining a relative address located just at the center of the relative address of the index I table. The binary search method thus divides a series of relative addresses (generally, a group of items) into two parts,
This is a search method that reaches (finds) a target relative address (item) by selecting one of them and further dividing the selected portion into two.

The index I table is accessed using the calculated relative address i, and the relative address i is accessed.
Index I (i) stored in the storage location having the index is read. This Index I (i) is set as k (step 55).

Subsequently, the index II table is accessed using IndexI (i) = k as a relative address, and the elements F1 (k), F2 (k), F3 (k) stored in the storage location having the relative address. ) Are read, and these are respectively set as p1, p2, and p3 (step 56).

Referring to FIG. 10, the tone code string / kanji code correspondence table is accessed using the third element F3 (k) = p3 read from the index II table as a relative address, and the relative address is obtained. The sound code Y0 (p3, C) stored in the storage location having is read out. FILT when the second mode is selected
ER = FEFC is set (FIG. 6, step 4
Five). This FILTER and the read sound code Y0
An AND logic operation with (p3, C) is performed. Also,
An AND logic operation is performed between the sound code x (C) specified by the sound code counter C of the input sound code string and FILTER. It is compared whether these two AND logical operation results are equal or which is larger (steps 63 and 64).

As described above, looking at the sound code string / kanji code correspondence table through the index I table, the sound code strings in the sound code string / kanji code correspondence table are arranged in ascending order of numerical values represented by them. Therefore, FILTER AND x (C) <FILTER AND Y0 (p3, C) Expression 1 holds that the input sound code x (C) is more than the read sound code Y0 (p3, C). This means that x (C) which is smaller, that is, to be searched for, is stored in a storage location whose relative address is smaller than Y0 (p3, C). To get closer to x (C), it is necessary to access a storage location with a smaller relative address. That is, it is necessary to access the upper half of the index I table. Then, i is substituted for the variable END (step 67), and the process returns to step 54 via step 53.

FILTER AND x (C)> FILTER AND Y0 (p3, C) In the case of Expression 2, i is assigned to the variable START (step
68), and similarly returns to step 54.

In this way, a sound code that matches the input sound code x (C) is searched in the sound code string / kanji code correspondence table according to the binary search method.

Finally, when FILTER AND x (C) = FILTER AND Y0 (p3, C) Expression 3 is satisfied, the storage location where the target sound code x (C) is stored is found in the correspondence table. .

If C = 1, then the second (C
= 2), the sound counter C is incremented in order to check whether the input sound code x (C) of the corresponding sound table matches the second sound code Y0 (p3, C) of the sound code string in the correspondence table. Step 66).

If Equation 3 is not satisfied in the incremented C, the search is performed again according to the binary search method according to which of Equations 1 and 2 is satisfied (Step 64). , 67, 68). Equation 3 holds when C = 1 means that the target sound code string must exist near Y0 (p3, C) found in the correspondence table, so that the binary search method is not used. , And its vicinity may be searched for. Further, a search can be performed using the element F1 (k) as described later.

The sound code Y0 (p3, C) for which Expression 3 is satisfied
As a result of incrementing the sound code counter C while finding all the sound codes, the search for all the sound codes in the input sound code sequence is completed, and when x (C) = φ is finally reached (step 61), It is checked whether or not Y0 (p3, C) = φ holds also for the sound codes of the sound code string in the table (step 69). Where Y
If ES is reached, a sound code string that matches the input sound code string has been found, and the process proceeds to the processing shown in FIG. If the sound code Y0 (p3, C) is not φ, the target sound code string exists at a position having a larger relative address than the storage location, so that the search is further continued for a storage location larger than the relative address. (Step 70). However, since a sound code string including all the input sound code strings was found (this is a case in which “Chinese” was found in an attempt to find “China”), the target sound was found in the immediate vicinity. Since there should be a code string, if the relative address in the correspondence table is incremented one by one, a target sound code string should be found immediately.

Further, when the read sound code Y0 (p3, C) has become φ while x (C) has not yet become φ (while searching for “China”, Is found) (step 62), the process proceeds to search for a storage location with a smaller relative address (step 65). In this case, too, the target character code string must exist immediately near the memory location that has been reached.
By decrementing the relative address of the correspondence table one by one, a target character code string can be quickly found.

Here, the meaning of FILTER = FEFC will be described with reference to FIG. For example, "Zhon
When the AND of FILTER = FEFC with any of the one to four voice codes having the vocal rhythm “g” is calculated, it is converted to a sound code representing the vocal rhythm “Zhong” ignoring the tone. Therefore, when a kanji search is performed in the second mode, all kanji characters having the same vocal rhythm are found regardless of the tone. This is shown in FIG. Regardless of whether the input includes a tone such as “Zhong1 Guo2” or the input without a tone such as “Zhong Guo”, the input sound code is filtered to obtain the expression “3” for the word “China”. Holds (at both C = 1 and C = 2), and this word is found. This means passing through the filters in the processing of steps 63 and 64.

In the first mode, it is not necessary to use a filter, and it is sufficient to directly compare the input sound code with the sound code in the correspondence table, but FILTER = F
If FFF is used, steps 63 and 64, and equations 1 to 3 can be used as they are.

When a sound code string that matches the input sound code string (through a filter) is found (step 69), referring to FIG. 11, the variable find is incremented, and the storage location of the kanji code buffer is specified. The kanji code KA (p.
3) is read from the sound code string / kanji code correspondence table and stored in the storage location specified by the variable find in the kanji code buffer (step 71).

Subsequently, using the second element F2 (k) of the index II table, the element F3 (k) of another storage location of the index II table that points to the sound code string of the corresponding table having the same pronunciation. ), The kanji code having the same pronunciation stored therein is stored in the kanji code buffer after specifying the new storage location by incrementing the variable find (step 73). ). The kanji codes having the same pronunciation linked by the element F2 (k) are sequentially read out and stored in the kanji code buffer (repeated steps 73 and 74). As a result, "chu"
Will be found as kanji candidates.

If F2 (k) = p2 = φ (step
72), a sound code string including and longer than the input sound code string is searched using the element F1 (k) = p1. The kanji code accessed by element F3 (k) stored in the other (second) storage location in the index II table pointed to by element F1 (k) is read from the correspondence table and the variable find Is incremented, and stored in the kanji code buffer (step 76). Element F1 (k) of second storage location
In the same way, if there are still other storage locations linked by
The kanji code accessed by (k) is read and stored in the kanji code buffer (steps 75 to 7).
7). If F1 (k) = p1 = φ, all the processing ends (step 75).

Further, by repeating the binary search,
If + 1 ≧ END, the kanji search process is terminated assuming that a sound code string corresponding to the input sound code string has not been found (step 53).

FIG. 17 shows a hardware architecture for realizing the above-described processing, and shows a configuration example of the kanji search processing device 17.

The sound code string / kanji code correspondence table 81 is the same as that shown in FIG. 12 or FIG. The search / readout circuit 82 reads out a sound code sequence from the correspondence table sequentially or by referring to the input sound code sequence. Data FFFF is stored in the filter register 83A,
Data FEFC is stored in the register 83B. In response to the conversion mode selection by the conversion mode key 24, the multiplexer 84 selects the register 83A in the first mode and the register 83B in the second mode, and stores the filter data stored therein in A.
It is given to ND circuits 85 and 86.

The input sound code string is stored in the AND circuit 85.
The D code 86 is provided with the sound code string read from the correspondence table 81, respectively. The input sound code string and the sound code string are filtered in the AND circuits 85 and 86, respectively, and their output data are compared in the comparison circuit 87. The comparison circuit 87 outputs a match signal only when the two input data match. The gate 88 is enabled by the coincidence signal. The match signal is supplied to the search / readout circuit 82. The search / readout circuit 82 reads out all the kanji codes corresponding to the matched sound codes from the correspondence table and supplies them to the gate 88. It will be stored in the kanji code buffer 89.

In the above embodiment, the first mode and the second mode
There is a mode, and the kanji search including the tone and ignoring the tone is performed using the filters FFFF and FEFC. However, the kanji search in other modes using the other filters may be performed. it can. For example, using a filter of 00FC in hexadecimal notation,
It will be possible to search for all kanji that have a phonetic code that matches the input voice code and the initial.

The input device is not limited to a keyboard, but may be a device for inputting pronunciation itself. In this case, a speech recognition device that converts an electric signal representing pronunciation into a corresponding sound code will be used.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an electrical configuration of a Chinese phonetic transcription / kanji conversion device.

FIG. 2 is a block diagram illustrating a hardware configuration of a main part of a Chinese phonetic transcription / kanji conversion device or a configuration from a functional viewpoint.

FIG. 3 shows an example of a Pin Yin / sound code conversion table.

FIG. 4 shows an example of a Zhu Yin / sound code conversion table.

FIG. 5A shows a data format of a sound code, and FIG. 5B shows a code representing a tone.

FIG. 6 is a flowchart showing a procedure of input and edit processing.

FIG. 7 shows how key input data is stored in a key data buffer.

FIG. 8 shows how sound codes are stored in a sound code string buffer.

FIG. 9 is a flowchart illustrating a kanji search processing procedure.

FIG. 10 is a flowchart showing a kanji search processing procedure.

FIG. 11 is a flowchart showing a kanji search processing procedure.

FIG. 12 shows the structure of a dictionary.

FIG. 13 shows an example of a sound code string / kanji code correspondence table.

FIG. 14 shows a Kanji code buffer.

FIG. 15 shows how a sound code is converted by filtering.

FIG. 16 shows a state of a kanji search using a filter.

FIG. 17 is a block diagram showing a hardware architecture of a kanji search device or focusing on a function of a kanji search process.

[Explanation of symbols]

 10 Computer 12 Display controller 13 Character generator 14 Display device 15 Switching circuit 16 Editing processor 17 Kanji search processor 20 Keyboard 21 Pin Yin key 22 Zhu Yin key 23 Input mode key 24 Conversion mode key 25 Other function keys 30 Memory device 31 Pin Yin / Sound code 32 Zhu Yin / Sound code 33 Sound code / Pin Yin 34 Sound code / Zhu Yin 35 Dictionary 36 Data after conversion 81 Sound code string / Kanji code correspondence table 82 Search / read circuit 83A, 83B Filter • Register 84 Multiplexer 85, 86 AND circuit 87 Comparison circuit 88 Gate 89 Buffer

Claims (10)

[Claims] A converting means for converting input data representing Chinese pronunciation into a sound code corresponding to the pronunciation;
A dictionary in which a sound code and a kanji code indicating a kanji having a pronunciation represented by the sound code are stored in correspondence with each other, filtering means for masking one or more predetermined bits constituting the sound code, By filtering the obtained sound code and the sound code of the dictionary using the filtering means and comparing them with each other, a sound code matching the sound code obtained from the conversion means is searched in the dictionary, and And a control means for reading a kanji code corresponding to a sound code to be performed from the dictionary. 2. The apparatus according to claim 1, wherein said tone code includes a bit representing an initial, a bit representing a rhyme, and a bit representing a tone. 3. The apparatus according to claim 2, wherein said filtering means masks a bit representing an initial, a bit representing a rhyme, or a bit representing a tone. 4. The apparatus according to claim 1, wherein said filtering means includes a means for passing a sound code as it is. 5. The apparatus according to claim 1, wherein said conversion means is a speech recognition means for recognizing a pronunciation based on a speech input signal and outputting a sound code corresponding to the pronunciation. 6. The apparatus according to claim 1, further comprising a search mode selecting unit for selecting whether or not the filtering unit is used, or selecting one of the plurality of filtering units. 7. The apparatus according to claim 1, further comprising: means for converting the read kanji code into display data representing a kanji indicated by the kanji code; and a device for displaying the kanji based on the display data. Equipment. 8. The apparatus according to claim 7, further comprising designation input means for designating any one of the displayed kanji candidates, and a memory for storing a kanji code indicating the designated kanji. . 9. The conversion means for converting input data of one kanji into a sound code, wherein the dictionary stores a sound code string and a kanji code corresponding to a word consisting of one kanji or a plurality of kanji. The control means controls the conversion means to convert the input data into sound codes by dividing the input data into one kanji, and arranges the converted sound codes in word units to generate a sound code sequence. 2. The apparatus according to claim 1, wherein a kanji code corresponding to the sound code string is searched in the dictionary. 10. A dictionary in which a sound code and a kanji code indicating a kanji having a pronunciation represented by the sound code are stored in association with each other in advance, and data representing an input Chinese pronunciation is stored. , Converted to a sound code corresponding to the pronunciation, and the sound code obtained by the conversion,
The sound code in the dictionary is filtered by masking one or more bits constituting the sound code, and compared with each other, and a sound code matching the sound code obtained by the conversion is searched in the dictionary. And a Chinese phonetic notation / kanji conversion method for reading a kanji code corresponding to a matching sound code from the dictionary.