JPS5827529B2 - Kanji/kana mixed sentence creation device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention provides a kanji/kana mixed sentence creation device, in particular, for example, from a character string input as a kana character string, words in the middle of the character string are compared with words stored in a word dictionary. Then, the kanji/kana mixture is determined based on the grammatical connection relationship with the words before and/or after the extracted word and the so-called longest match method, and the above character string is converted into a kanji/kana mixture sentence. In the sentence creation device, when the words extracted according to the longest match method match more than a predetermined number of words, the word is determined and output without checking the grammatical connection relationship. The invention relates to a device for creating sentences containing mixed kanji and kana.

Recently, kana/kanji conversion devices are being developed that input kana character strings and convert them into sentences containing kanji/kana.

In a device for creating sentences containing kanji and kana using such a conversion device, it is necessary to extract words from the input kana character string and convert them to kanji6. Two effective methods are used: a method that examines the grammatical connections before and after the extracted word, and a so-called longest match method that extracts the longest match by comparing the words stored in a word dictionary. ing.

However, in conventional devices of this kind, the word processed first is considered to be correct, and if the following word does not have the correct grammatical connection, the following word is Even if a long number of words matches the stored word, it is rejected.

This method of the conventional device is effective as long as the above-mentioned human character string is inputted correctly, since the grammatical connections are maintained correctly.

However, if there is an input error when inputting the above-mentioned human character string, the relevant part may be incorrectly converted, and the conversion may continue based on grammatical connections, or the subsequent word conversion may be incorrect. Sometimes things become impossible.

The purpose of the present invention is to solve this problem, and on the condition that a match is made with a stored word exceeding a predetermined number of words, the matching word can be used without checking the grammatical connection. The purpose is to output the data as correct, and to be able to proceed with the process without compromising the principle of checking grammatical connections.

To this end, the kanji/kana mixed sentence creation device of the present invention is designed to ensure that, for input character strings in which at least the part to be converted into kanji is input as kana characters, the word in the first part of the character string corresponds to the word dictionary. The longest match method selects the word that has the longest match among the words that are matched with the grammatical connection relationship with the previous and/or subsequent word and the word stored in the word dictionary. The kanji/kana mixed sentence creation device for converting the input character string into a kanji/kana mixed sentence is provided with a word dictionary search unit and a grammatical information check circuit unit. When matching the words in the word dictionary with the word dictionary, the words are extracted by the longest match method, and words that match more than a predetermined number of words are output without being checked by the grammar information checking circuit, Further, the grammar information checking circuit section checks and outputs the words that match within the predetermined number of words.

This will be explained below with reference to the drawings.

Fig. 1 shows the structure of an embodiment of the present invention, Fig. 2 is an explanatory diagram explaining the grammatical connection relationships of words, Fig. 3 shows the structure of an embodiment of a word dictionary, and Fig. 4 shows the structure of an embodiment of the word dictionary. An example of a resulting input character string, FIG. 5 shows an example of processing performed by a conventional device, and FIG. 6 shows an example of processing according to the present invention.

In Figure 1, 1 is a character string input device, 2 is a buffer
3 is a register, 3 is a word dictionary search device, 4 is a word dictionary, 5 is a grammar information check circuit, and 6 is an output file.

Prior to explaining the present invention based on FIG. 1, the concept of the conventional device and problems with the conventional device will be explained.

It can be considered that the conventional device did not have the route indicated by diagonal lines in the configuration shown in FIG.

In Japanese sentences, as shown in Figure 2, it is determined whether each word with the part of speech shown in the left column follows the word with the part of speech shown in the upper horizontal row (the places marked with an asterisk are grammatical). Based on this, a word dictionary as shown in FIG. 3 is prepared as the word dictionary 4 shown in FIG. 1.

In other words, the kanji corresponding to the word with the heading "gin" is given as "gin" in the form of a kanji code, and below, for example, in response to the heading "ginkora", "ginkoraman", etc. "Bank", "Bankman" in the form of Kanji code
. . . are stored in the word dictionary 4.

Then, for each heading/kanji code correspondence, the number of words in the heading (heading length) and the number of words when converted to kanji (
Kanji code length) and an index (INDX) indicating what part of speech the word can follow and what part of speech the word can follow.

And if we now have the word ``Ginkora'' in Word Dictionary 4.
When searching for , the stored words ``silver'' and ``bank'' match, but the longest match method is used to extract the one with the longer heading length, ie, ``bank.''

Then, grammatical connections are checked using the index.

To explain this point with reference to FIG. 1, a kana character string is input by a character string input device 1 and set in a buffer register 2.

Then, the word dictionary search device 3 retrieves the characters (a t b t Ct d t e + f
y g y ”' ) to search the word dictionary 4 and find the words (a) (a, b), (a, b+c) stored in the dictionary 4.
), (a, b.

c, d), the word (a, b, c, d) is extracted as existing.

Then, the character string (e, 19g,
) are left and the grammatical connection relationship of the extracted words (a, b, c, d) is checked by the grammatical information checking circuit unit 5.

That is, the word immediately before the extracted word (X, Y, Z
) has been determined, the determined word (X,
The grammatical connection relationship between the words (Y, Z) and the extracted words (a, b, c, d) is checked.

And if it is grammatically satisfied, the word extracted above (
a, b, c, d) are determined as correct words and output to the output file 6.

If it is not grammatically correct, use the above words (a, b, c,
d), for example, the next longest matching word (a,
b, c) are extracted.

In the conventional apparatus, processing is performed as described above, and the route shown by diagonal lines in FIG. 1 does not exist.

For this reason, if the place that should have originally been entered as "Ginkou detsu ka ji do yokin kiha..." is mistakenly input as "Ginkou detsu ka ji do yokin kiha..." as shown in Figure 4, then This results in processing as shown in FIG.

In other words, even if the next word [-automatic teller machine] is extracted from the incorrectly inputted ``e'' based on the longest match method, the noun is Since it cannot be connected, the word "automatic teller machine" is rejected.

As a result, words connected to ``1e'' cannot be found, and ``ji'', ``do'', etc. are output without conversion.

Also, in some cases, if a word like "extension telephone" happens to exist after the above "e" instead of "automatic deposit machine" above, it will be processed as "It's a line that can't be used at banks..." It is possible.

In other words, an erroneous conversion occurs due to an input error.

In order to improve this point, in the case of the present invention, a diagonal route shown in Figure 1 is provided, and for example, if the word extracted by the longest match method exceeds 1-5 words,
Words are output as correct words without checking grammatical connections by a grammatical information checking circuit section 5.

Then, the grammar information checking circuit section 5 checks only words with a word count of "5" or less.

The configuration for this purpose is to check the word length of the longest matching word in the word dictionary search device 3 shown in FIG.
” It is sufficient to pass only the following to the grammar information checking circuit unit 5.

By doing this, even when converting a character string with an input error as shown in Figure 4, the word "automatic deposit machine" is determined as the correct word as shown in Figure 6, and henceforth "is - 1... will be converted.

In other words, "Use/U" that should have been entered is "Use".
Even if it is entered incorrectly, the conversion process will return to normal once a word with a long number of words such as "automatic teller machine" is extracted.

As explained above, according to the present invention, the chain of erroneous conversions caused by input errors can be broken and correct processing can be performed without fundamentally breaking the principles of the longest match method and grammatical connection relation checking. It is possible to return to.

[Brief explanation of drawings]

Fig. 1 shows the structure of an embodiment of the present invention, Fig. 2 is an explanatory diagram explaining the grammatical connection relationships of words, Fig. 3 shows the structure of an embodiment of a word dictionary, and Fig. 4 shows the structure of an embodiment of the word dictionary. An example of a resulting input character string, FIG. 5 shows an example of processing performed by a conventional device, and FIG. 6 shows an example of processing according to the present invention. In the figure, 1 is a string input device, 2 is a buffer register,
3 is a word dictionary search device, 4 is a word dictionary, 5 is a grammar information check circuit, and 6 is an output file.

Claims (1)

[Claims] 1 For input character strings in which at least the part to be converted into kanji is input as kana characters, the words in the character string are matched with the word dictionary and the grammatical connections with the previous and/or subsequent words are determined. It is determined based on the longest match method, which selects the word with the longest match among the words stored in the word dictionary, and converts the input character string into a kanji/kana mixed sentence. Kanji to be converted into
The kana-mixed sentence creation device includes a word dictionary search unit and a grammar information check circuit unit, and extracts words by the longest match method when matching words in the input character string with the word dictionary; Words that match more than the predetermined number of words are output without being checked by the grammar information check circuit, and words that match less than the predetermined number of words are checked by the grammar information check circuit. A device for creating sentences containing mixed kanji and kana characters.