JPH01228058A - Kana/kanji converter - Google Patents

Abstract

PURPOSE:To raise a conversion hitting rate by investigating the meaning relation of a compound word and determining displaying priority with a Chinese word meaning classification and the meaning classification to cooccur by the affix. CONSTITUTION:A Chinese word meaning classifying dictionary 4 and an affix cooccurring dictionary 5 are connected to a KANA (Japanese syllabary)/KANJI (Chinese character) converter besides a Chinese word dictionary 3. In the Chinese word meaning classifying dictionary 4, an action, organization, somatology, qualification, etc., and information classifying the meaning shown by them are stored concerning a Chinese word, etc. In the affix cooccurring dictionary 5, the affix, the meaning classification to cooccur at the affix and the evaluation value for the meaning classification are stored like an end-action, a store-action and merchandise. Concerning the inputted sentence, only the word able to be connected with a preceding word is remained at a candidate storing part 9 by a connecting detecting part 14, only the connection having a meaning is extracted by a compound word processing part 15, the evaluation value is calculated by an evaluation value calculating part 16 and the word is selected by a word selecting part 12.

Description

DETAILED DESCRIPTION OF THE INVENTION Technical Field The present invention relates to a kana-kanji conversion device such as a Japanese word processor.

BACKGROUND TECHNOLOGY In general, this type of kana-kanji conversion device displays a plurality of possible conversion candidates for an input kana character string and allows the user to select one.

At this time, in the case of a compound word whose constituent elements are a prefix or a suffix, conversion candidates are generated by checking whether the compound word itself is registered or by simply checking the connections. If the compound words themselves were to be registered, that is, if only the correct combinations were to be registered in the dictionary, the number of registered words would increase enormously, and it would be nearly impossible to register all of them, and it would be inefficient.

On the other hand, according to the method of generating possible conversion candidates by examining only connections, there is no need to register each compound word itself.
You can save dictionary storage space. However, when candidates are generated solely based on connections, conversion candidates for compound words that are combinations that are not semantically correct are often output. As a result, the operator may find it time-consuming to select, or incorrect words may be preferentially selected and written into the document, which is cumbersome and has poor operability.

Purpose The present invention has been made in view of the above points, and it is possible to improve operability without increasing the storage capacity of a dictionary and without outputting semantically incorrect words as compound word conversion candidates. The purpose is to obtain a kana-kanji conversion device that can be used.

Structure In order to achieve the above-mentioned object, the present invention provides a Chinese semantic classification dictionary that stores semantic classifications of Chinese words and Chinese equivalent words, and a semantic classification and semantic classification of prefixes, suffixes, and affix equivalent words that co-occur with these. Using an affix co-occurrence dictionary that stores evaluation values for each, and these Chinese meaning classification dictionaries and affix co-occurrence dictionaries, we examine the internal semantic relationships of compound words whose constituent elements are prefixes and suffixes, and investigate the semantic relationships between them. and compound word processing means that displays and selects correct words preferentially as conversion candidates based on the conversion information.

Hereinafter, one embodiment of the present invention will be described based on the drawings. First, schematically, an input device 1 such as a keyboard, a CRT
an output device 2 such as, a word dictionary 3 that extracts the notation from the input pronunciation, a Chinese semantic classification dictionary 4 and an affix co-occurrence dictionary to which semantic information is added in order to perform compound word processing, which is a special feature of this embodiment. 5 is connected to the kana-kanji conversion device main body 6 to configure the system.

The word dictionary 3 stores information such as pronunciation, part of speech, frequency, priority order to be output in the case of the same pronunciation, and notation, as shown in FIG. 2, for example.

In addition, the Kanji meaning classification dictionary s4 includes Kanji words that are formed by combining two or more Kanji characters, katakana, etc. that make up Kanji words together with prefixes and suffixes (i.e., words that make up compound words, and are called Kanji equivalent words). ), and their meaning classifications are stored as shown in FIG.

The semantic relationships of compound words are checked by the semantic classification in the Chinese semantic classification dictionary 4 described in Section 2.

As shown in FIG. 4, the affix co-occurrence dictionary 5 stores combinations of notations of affixes, semantic categories that co-occur with the affixes, and evaluation values for each semantic category. At this time, the evaluation value for each semantic classification is set higher as it is semantically more correct and frequently used.

Furthermore, the kana-kanji conversion device main body 6 can use these dictionaries 3 and 4.
．． 5, etc., to actually perform kana-kanji conversion processing, and has built-in memories consisting of a Chinese language storage section 7, an affix storage section 8, and a candidate storage section 9.

Here, when the previous word is a prefix, the Chinese language storage section 7 is a part that extracts and stores two or more Chinese characters in the conversion candidate from the candidate storage section 9. FIG. 5 shows an example. That is,
This is an area for matching with the meaning classifications stored in the Chinese meaning classification dictionary 4. Further, the affix storage section 8 is a section that extracts and stores only affixes, which are prefixes or suffixes, from among conversion candidates from the candidate storage section 9. FIG. 6 shows an example. That is, this is an area for matching with co-occurrence meaning classifications (including evaluation values) stored in the affix co-occurrence dictionary 5. Further, the candidate storage section 9 stores words that can be candidates for kanji conversion after dictionary searches and connection tests. An example thereof is shown in FIG. Words that are not subject to compound word processing are also held in this candidate storage section 9.

In the kana-kanji conversion device main body 6, the dictionary search unit 10
, a word evaluation section 11 , a word selection section 12 , and a confirmation processing section 13 , which perform kana-kanji conversion processing. The present embodiment is particularly characterized by the processing in the word evaluation unit 11 after the search by the dictionary search unit 10. This word evaluation section 11 has a processing section consisting of a connection test section 14, a compound word processing section 15 serving as a compound word processing means, an evaluation value calculation section 16, etc., and executes processing according to the flowchart shown in FIG. 17 is a backtrack.

Briefly, after the dictionary search section 10 searches the dictionary, the connection verification section 14 checks the connection with the previous word, and only those that can be connected are held in the candidate storage section 9. This candidate storage section 9
The compound word processing unit 15 performs compound word processing on the words held in . After this compound word processing, an evaluation value calculation unit 16 calculates an evaluation value for each conversion candidate. After the processing by the word evaluation section 11, the processing is advanced to the word selection section 12,
Words are selected from these conversion candidates and evaluation values and passed to the confirmation processing section 13.

The processing of this embodiment will now be described using a specific example while referring to the flowchart of FIG.

Example 1) Assume that there is an input character string [Unkouyamishiyukkomonowa], and the character string "Mi" at the beginning of the sentence has been determined.

First, if you search the dictionary after [Mi], the pronunciation of ``koukou'' will be searched for ``filial piety,'' ``high school,''``oral,''``navigation,'' and the pronunciation of ``koukoya'' will be searched for ``kookoji,'' and these will be searched for in the candidate storage section 9. Register. Then, the process moves to compound word processing, but since these candidates have no suffix and the immediately preceding prefix is "mi", Chinese words with two or more characters are extracted into the Chinese language storage section 7.

The Chinese language storage section 7 in FIG. 5 shows the state in this case.

Then, the remaining candidates that were not extracted are held in the candidate storage unit 9, but in this case, all the candidates mentioned above are extracted,
There are no remaining candidates, and the candidate storage section 9 becomes empty. Next, the prefix is matched with the meaning classification of the Chinese word extracted into the Chinese word storage section 7. In this example, the prefix “Mi”
As shown in FIG. 4, the co-occurrence meaning classification of is "action", so in the above Chinese words extracted into the Chinese language storage unit 7, "action" is
Only "navigation" with the semantic classification is matched.

Therefore, we extracted the Chinese word for "navigation" and extracted the candidate "navigation".
"Navigation" is inserted at the beginning of the candidate storage section 9.

Since there is a word in the candidate storage section 9, the compound word processing is completed and the process is moved to the evaluation value calculation section 16 with the candidate "navigation" and its evaluation value "3". In this way, the correct compound word "unsailed" is selected as a candidate.

Example 2) Assume that there is an input character string ``When you can't get a passing grade'' and that the beginning of the sentence, ``r passed,'' has been confirmed.

First of all, when you search the dictionary after "pass", the reading is "ten".
``Ten'', ``Ten'', ``Ten'', ``Ten'', ``Ten'', and ``Ten'' are obtained, and ``Tenga'' is searched from the pronunciation ``Tenga'', and these are registered in the candidate storage section 9. FIG. 7 shows an example of the state of the candidate storage section 9 at this time. Then, we move on to compound word processing, but since these candidates have suffixes,
A suffix is extracted from the candidate storage section 9 and registered in the affix storage section 8.

FIG. 6 shows the state of the affix storage section 8 at this time. and,
The remaining candidate "Noga" which is not a suffix is kept in the candidate storage section 9. Next, the semantic classification of the immediately preceding Chinese word ``pass'' and each extracted suffix candidate is performed. In this case, from the co-occurring semantic categories, the items that can be matched with "pass" are "1 store", "exhibition", and 1 item, and these are extracted from the affix storage unit 8. There are multiple candidates, so arrange them in descending order of evaluation value (in this example, 1 point).
"Exhibition") and are pushed to the beginning of the candidate storage section 9 in this order. Therefore, after this processing, one item is stored in the candidate storage unit 9 in the order of ``Shop'', ``Exhibition'', and ``Elegance''.

In this manner, the process ends with the candidates stored in the candidate storage unit 9, and the process proceeds to the evaluation value calculation unit 16. In this case as well, the candidates are those that are semantically correct, such as "passing score" and "passing store" and "passing exhibition." In other words, words that are semantically incorrect, such as "passing turn" and "passing tool", are not selected as compound word candidates.

According to this embodiment, not only the connections but also the internal meanings of compound words are examined and only words that are semantically correct are output as conversion candidates, so that the operator can select the desired compound word candidate. It has a high probability, is not confused by semantically incorrect information, and is easy to operate. For this reason, instead of registering compound words, it is sufficient to store information for each Chinese word and information for each affix in Dictionary 4.5, which requires less storage capacity. be.

Note that the system of this embodiment can be applied to a proofreading support system and a machine translation system by using the same dictionary and matching method even if the input character string is a mixture of kanji and kana.

Effects As described above, the present invention uses a Chinese meaning classification dictionary that stores the meaning classifications of Chinese words, etc., and an affix co-occurrence dictionary that stores the meaning classifications of affixes that co-occur with affixes and their evaluation values. The word processing means also examines the internal semantic relationships of compound words whose constituent elements are prefixes and suffixes, and based on these semantic relationships, the correct words are displayed preferentially as conversion candidates and selected. Unlike the case where candidates are generated based on relationships, semantically incorrect words are not selected as candidates, and only semantically correct words can be selected as compound word candidates, improving operator operability during selection. For this purpose, it is sufficient to provide a Chinese meaning classification dictionary and an affix co-occurrence dictionary, and the storage capacity does not increase as in the case of storing the words in the form of compound words.

[Brief explanation of the drawing]

The drawings show one embodiment of the present invention, and FIG. 1 is a block diagram;
Figure 2 is an explanatory diagram of the word dictionary, Figure 3 is an explanatory diagram of the Chinese meaning classification dictionary, Figure 4 is an explanatory diagram of the affix co-occurrence dictionary, Figure 5 is an explanatory diagram of the Chinese word storage unit, and Figure 6 is an explanatory diagram of the affix. Explanatory diagram of storage unit, 7th
The figure is an explanatory diagram of the candidate storage section, and FIG. 8 is a flowchart showing compound word processing. 4... Chinese meaning classification dictionary, 5... Affix co-occurrence dictionary, 1
5... Compound word processing means

Claims (1)

[Claims] A Chinese semantic classification dictionary that stores the semantic classifications of Chinese words and Chinese equivalent words; and an affix co-occurrence dictionary that stores the semantic classifications that co-occur with prefixes, suffixes, and affix equivalent words and evaluation values for each semantic classification. , uses these Chinese semantic classification dictionaries and affix co-occurrence dictionaries to examine the internal semantic relationships of compound words whose constituent elements are prefixes and suffixes, and based on these semantic relationships, correct words are displayed preferentially as conversion candidates. A kana-kanji conversion device comprising a compound word processing means for selecting.