JPS588379A - Kana (japanese syllabary)-kanji (chinese character) converting system - Google Patents

Abstract

PURPOSE:To quickly obtain the result of desired connection and to perform the converting processing at high speed, by designating a conversion candidate even for a word short in the length of coincidence with an inputted Kana train, when the margin of connection with a character or a word connected in front is high. CONSTITUTION:A front connection character is read out to a character confirming section 3 from an input character buffer 1 with a conversion control section 2 of a Kana character conversion system, and the type of character of front-connected character is designated. In this case, when the type of character is other than Kana, the character type code and when Kane, the word of conversion condidate, corresponding to the part of end of the Kana are read out from conversion condidate buffer 6, the part of speech is recognized at a part of speech recognizing section 4 and the type code is inputted to a connection margin recognizing section 5. Words are sequentially given to the section 4 from word groups with different length of coincidence corresponding to the Kana train of the position of the front connected character from the control section 2 to recognize the part of the speech. The conversion candidate is stored in the buffer 6, and Kana next to the candidate is retrieved at a sentence retrieval section 7, and the result of connection desired is applied to the character output buffer 8 quickly.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a kana-kanji conversion method useful for Japanese document creation devices, etc., and aims to perform conversion at high speed.

Conventionally, when converting a sentence-based kana string into a kanji character, the most important consideration was the length of the match with the converted kana string, as seen in the longest match method. For this reason,
If a word with a short matching length is correct, the word with a long matching length is first set as a conversion candidate, and the correct answer cannot be obtained unless the word is eliminated through a grammar check or the like.

Figure 4 shows an example. A is the input character string to be converted to kanji, B is the academic language search result, and C is the conversion example. The correct kanji "kai" is a conversion candidate because the match length is short. As a result, the ranking will be lower.

The present invention aims to completely eliminate such defects, and the following will be explained below.
An embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of the kana-kanji conversion method of the present invention. In FIG. 1, 1 is an input character buffer, 2 is a conversion control unit, 3 is a character type recognition unit, 4 is a word part-of-speech recognition unit, 6 is a likelihood recognition unit for character types, parts of speech, and connections between parts of speech, and 6 is a conversion unit. Candidate buffer, 7 is clause search section, 8
is the output character buffer.

FIG. 2 is a diagram showing an example of an input character string, an example of a corresponding word group, and an example of a conversion result for explaining the present invention. In the figure, A is an input character string. , B is the word search result based on the position of "force", and C is the conversion result.

FIG. 3 is an example of a connection likelihood table used by the connection likelihood recognition unit 6 for likelihood recognition, in which the part-of-speech code and character type code of the front and rear words are arranged in rows and columns, respectively. The likelihood of connection between the corresponding part-of-speech code and character type code is set at the intersection. In this example, large numbers are given to locations with high likelihood.

The operation of each part shown in FIG. 1 will be explained using the examples shown in FIGS. 2 and 3.

As shown in Figure 2, the prefix of "power" is the number "8".
, and as a result of a word search based on the position of "force", the second
Assume that the words in part B of the diagram are obtained.

The conversion control unit 2 reads the character prefix “8” from the input character buffer 1 to the character type recognition unit 3 to recognize the character type, and converts the character type to kana, which ends with that kana, as in the current example. The corresponding conversion candidate word is read from the conversion candidate buffer 6, the part of speech is recognized by the part of speech recognition section 4, and the part of speech code is passed to the connection likelihood recognition section 6. On the other hand, conversion control section 2
The part-of-speech recognition unit 4 sequentially selects words from a group of words that include words with different matching lengths that can correspond to the kana sequence starting from the position of “power”.
The part-of-speech code is passed to the connection likelihood recognition unit 6. The connection likelihood recognition unit 6 uses the conversion control unit 2 based on a connection likelihood table between the front part-of-speech code or character type and the rear part-of-speech code or character type, as shown in FIG. 3, for example.
Under the control of
The likelihood of connection with the character type code is recognized and passed to the conversion control unit 2. Based on the connection likelihood, the conversion control unit 2 converts the word group from the word group as shown in FIG. ``Kaisaku,''``Kai.''

・・・・One by one in the order of −, therefore, first store “Kai” as a conversion candidate in the conversion candidate buffer 6, and then store it from the next kana “su” of the kana part to which “Kai” corresponds. Perform a word search. The conversion control unit 2 further deletes the prefix word from the conversion candidate buffer 6 if, for example, there is no matching word as a result of the word search or there is no single order connected to the prefix word or prefix character, An operation of newly storing the next ranking word in the conversion candidate buffer 6 is performed. This kind of behavior
Until the conversion candidate group corresponds to the entire corresponding kana county,
Do it repeatedly. The second line corresponds to the underlined part in Figure 2 (8).
A group of conversion candidates as shown in the underlined part of the figure (q) are being searched.

At this point, the conversion control unit 2 transfers the conversion candidate group to the output character buffer 77 and ends the operation.

In this way, if a word with a high connection likelihood is selected and the conversion is successful, wasteful processing for words with a lower connection likelihood can be omitted.

As is clear from the above explanation, according to the present invention, even if the length of the match with the input kana string is relatively short, if the likelihood of connection with the preceding character or word is high, it is first selected as a conversion candidate. Therefore, the desired conversion result can be obtained quickly, and the conversion time can be shortened compared to the conventional method of sequentially treating words with the longest match length as conversion candidates.

Note that the connection likelihood shown in Figure 3 is merely an example, and the classification of parts of speech and character types, as well as the numerical value given as the likelihood, are not limited to 9 and may be set as appropriate. Needless to say.

[Brief explanation of the drawing]

Figure 1 is a block diagram showing a kana-kanji conversion method in an embodiment of the present invention, Figure 2 is a diagram explaining the conversion process, Figure 3 is a diagram used for explanation. A diagram showing an example of connection likelihood, Figure 4 (8), (B), (Q
1 is a diagram showing an example of conversion in a conventional conversion method. 1... Input character buffer/buffer, 2... Conversion control section, 3... Character recognition section, 4...
Part of speech recognition, fixed part, 6... Connection likelihood recognition part, 6...
... Conversion candidate buffer, 7 ... Clause search unit, 8 ... Output character buffer. Name of agent: Patent attorney Toshio Nakao and 1 other person11
Diagram (A) Type 8 swords = Σ4J to chini tjikora
! (c) It's the 8th line 821" (A) ・Tai 8114 Sakura Matsuri? (8)
Difficulty Particle Ward I Month Name iq Keishi Noun 0 Day Noun Same Particle [Mouth Suffix (C) 8th 4th Festival〆

Claims (1)

[Claims] ′ It is fully equipped with a means to recognize the character type, a means to recognize the part of speech of a word, and a means to recognize the likelihood of connection between the character type, the part of speech of the word, and the parts of speech of the word, and corresponds to a part of the kana sequence. From a vine word group, each word in the word group, the word to which the kana part is converted if the part is preceded by a kana part, and the character if the part is preceded by a non-kana character. A kana-kanji conversion method characterized in that the likelihood of a connection is determined based on the part of speech of a word and the character type of a non-kana character, and words with the highest likelihood are selected as conversion candidates.