JPH01237877A - Kanji conversion system - Google Patents

Abstract

PURPOSE:To heighten the learning effect of a compound word by registering the compound word on a provisional dictionary when the compound word is selected, and performing following learning on the provisional dictionary. CONSTITUTION:A KANA (Japanese syllabary) character string inputted from an input part 1 is delivered to a word retrieval part 2. The word retrieval part 2 takes out a word and the compound word as many as possible to referring to the provisional dictionary 4 and a dictionary 3. A priority order by learning is attached on a taken out word, however, at this time, a high priority order is attached on the word taken out of the provisional dictionary 4. A compound word generating part 5 assembles the compound word from each word delivered from the word retrieval part 2 when necessary, and after that, delivers it to a user selection part 7 with the word delivered from the word retrieval part 2. When the compound word is selected, it is registered if it is not registered on the provisional dictionary 4, and the highest priority order is attached on the word.

Description

DETAILED DESCRIPTION OF THE INVENTION (Technical Field of the Invention) The present invention relates to a kanji conversion method for efficiently learning compound words.

[Prior art and its problems]

When converting readings such as kana to kanji or sentences containing kanji and kana (hereinafter collectively referred to as kanji conversion), if there are homophones, multiple kanji conversion candidates can be created. In such cases, conversion candidates are generally presented in an order to the user for selection. This display order follows the priority order of homophones written in the dictionary in which each word is stored, and the priority order is learned so that the one selected most recently by the user has a higher priority. As a result, the homophone and allograph word once selected by the user is then displayed as a high priority, such as the first candidate. Considering the fact that one particular word among homophones is often used in the same sentence, this type of learning greatly improves the efficiency of kanji input. This is explained, for example, on page 70 of the Nikkei Electronics Book "Japanese Processing" published December 12, 1983.

The number of words that can be held in a dictionary is limited by the dictionary size and the speed of access to the dictionary.

Therefore, words that cannot be stored in the dictionary are synthesized using an algorithm from words stored in the dictionary. In other words, by cutting out the part of the input pronunciation that partially matches the pronunciation of the word in the dictionary, the input pronunciation is broken down into constituent elements, and the conversion result corresponding to each element is converted according to a predetermined rule. Some methods are being used, such as combining them. Since this itself is well known to those skilled in the art, further explanation will be omitted herein.

When this compound word is selected from among other homophones and allographs, there is no corresponding word in the dictionary, so the above-mentioned learning cannot be applied as is. Therefore, by learning each word that was an element of the compound word and raising the priority of these words, we tried to improve the conversion efficiency when the same compound word appeared next time.

However, this conventional method suffers from the disadvantage that learning of the entire compound word is essentially not performed, and the usability cannot be said to be sufficiently good.

As an example of this inconvenience, consider the case where the words "report meeting" and "top floor" are generated as compound words of "report" and "kai" and "mogami" and "floor."

After converting ``report meeting'' into kanji, the priority of the kanji ``kai'' among the homophones corresponding to the reading kaiJ increases, and after converting ``the top floor'' into kanji, it becomes ``floor.'' Kanji will have higher priority.

Therefore, if a document in which the words "report meeting" and "top floor" both appear many times is input using a device based on such a kanji conversion method, the learning effect of the reading of "kai" will result in the words "reporting floor" and " A conversion candidate named ``Mogamikai'' frequently appears in the first rank, and learning results in a decrease in conversion efficiency.

[Problem of invention]

The object of the present invention is to solve the problems of the prior art described above,
An object of the present invention is to provide a kanji conversion method that allows sufficient learning of compound words.

[Summary of the invention]

According to an embodiment of the present invention, when a compound word is selected, the compound word is registered in a provisional dictionary, and subsequent learning is performed on this provisional dictionary, thereby increasing the learning effect of compound words. A Kanji conversion method is provided.

[Embodiments of the invention]

FIG. 1 is a block diagram showing the configuration of a kana-kanji conversion system constructed by applying an embodiment of the kanji conversion method of the present invention.

In the figure, a kana character string input from an input section 1 is passed to a word search section 2. The word search unit 2 refers to the provisional dictionary 4 and the dictionary 3 and extracts as many words and compound words (hereinafter collectively referred to as words) already registered in the provisional dictionary from this kana character string as much as possible.

Here, Dictionary 3 contains Kana-Kanji conversion data that was registered from the beginning by the creator of this Kana-Kanji conversion system, as well as Kana-Kanji conversion data that has been additionally registered by users of this system, and also contains words with homophones and allographs. Priority orders set in advance are also registered. The provisional dictionary 4 will be described in detail below, but compound words and words selected by the user during kanji conversion are automatically registered here with priorities determined by learning.

The words retrieved by the word search section 2 are passed to the compound word generation section 5. If there are homophones and allographs as a result of retrieval by the word search unit 2, they are given in priority order. At this time, words taken out from the temporary dictionary 4 are first given a high priority. After the words from the provisional dictionary 4 are exhausted, the words from the dictionary 3 are assigned a priority order according to the order registered in the dictionary 3 as a lower order than the words from the provisional dictionary 4 and handed over.

The compound word generation section 5 assembles a compound word from each word passed from the word search section 2, if necessary, and passes it to the user selection section 7 together with the word passed from the word search section 2. When this compound word is assembled, the compound word has a lower priority than any word from the word search unit 2. Furthermore, when a plurality of compound words are generated by the compound word generation unit 5, how to determine the mutual priority among them is the priority of each element of the compound word, the number of elements making up the compound word, the connection between each element. This can be done based on grammatical rules such as degrees and various other factors. Since this determination method itself is not directly related to the gist of the present invention, it will not be further explained here.

The user selection unit 7 selects the word (
(originally retrieved from the dictionary 3, provisional dictionary 4 by the word search unit 2) and compound words (if any) by presenting them to the user according to their associated priority order. The user is forced to choose what he or she intends to do by using means that are not available. The specific implementation methods of the presentation method and selection method used here are well known, so they will not be explained further.

The homonyms selected by the user in this manner are passed to the output section 8. If this homonym is not the first priority, it is also passed to the provisional dictionary registration unit 6 at the same time. The provisional dictionary registration unit 6 performs registration processing of this homophone and allograph word in the provisional dictionary 4. In this process, if the same homophone and allograph that has been passed is already registered in the provisional dictionary 4, its priority is raised to the first priority.

If it has not been registered yet, register that homophone in the provisional dictionary 4, and if a homophone with the same pronunciation and other spellings has already been registered, change the priority of the newly registered homophone. Make it the top.

By performing such a registration process, even a compound word that does not exist in the dictionary 3 is registered in the temporary dictionary 4 in a synthesized form, and is studied there in the same way as a normal word.

Words in the dictionary 3 are also registered in the temporary dictionary 4 if they are used. Therefore, if the same reading is given, the same word will be retrieved from both the dictionary 3 and the provisional dictionary 4, which may confuse the user. In order to avoid this, for example, the word search section 2 or the compound word generation section 5 should detect such duplication and suppress the second duplication.

Note that the learning results and new compound words registered in the provisional dictionary 4 may be additionally registered in the dictionary 3 later. In addition, if multiple users take turns using this system, or if there is a possibility that multiple users will use the system at the same time, Dictionary 3 can be used as a common dictionary, and Temporary Dictionary 4 can be used for each. It can also be configured as a dictionary for each person. Further, a plurality of provisional dictionaries 4 may be prepared depending on the type of document to be inputted, field, etc.

In addition, in order to clarify the essence of the present invention, in the block diagram shown in FIG. 1, components that belong to the prior art and the flow of data and control between these components are shown in a simplified manner. However, it goes without saying that it is extremely easy for those skilled in the art to refer to this and implement the present invention.

Note that the present invention can be realized in various forms other than the embodiments described above.

For example, is it possible to enter a fairly long reading without any breaks?As a compound word pattern that frequently appears in kanji conversion, there are cases where a number is inserted in the middle, such as ``dai 12 kai J''.Even in the case of compound words like this, , for the reason explained at the beginning, it is inconvenient to learn only each word that makes up a compound word, so it is effective to apply the present invention.However, in this case, it is effective to apply the present invention. Furthermore, it is desirable that the learning effect also be applied to items in which numbers are replaced with other numbers. Therefore,
In this case, instead of registering the numeric part as is in the provisional dictionary, for example, ``Dai*kaiJ'', a symbol that indicates a match with any character string consisting of digits (here * is the symbol that represents it) is used. It goes without saying that this method can be applied to situations other than numbers where it would be more convenient to apply such a "wildcard" rule.

In addition, in the example shown above, as a learning algorithm,
The method uses a method that raises the priority of the word that was used immediately before, and lowers the priority of other words with the same reading, but the present invention can be used in conjunction with other algorithms. Can be done.

Further, only compound words and their learning results may be registered in the provisional dictionary 4, and the results of learning words already registered in the dictionary 3 may be written in a location other than the provisional dictionary 4, such as the dictionary 3.

Further, in the embodiment shown above, a kana character string is given to the input unit 1 as the reading, but any input expressing the reading other than kana, such as Roman characters, may be given. Furthermore, the given pronunciation may be converted into another expression format, such as Romaji → Kana → Kanji (a mixture of Kanji and Kana), and then the dictionary may be searched.

〔Effect of the invention〕

As explained in detail above, according to the present invention, the same learning effect can be obtained for compound words as for words that are registered in the dictionary from the beginning, so the conversion efficiency can be greatly improved as the words are used more effectively. system can be realized.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a kana-kanji conversion system constructed using an embodiment of the kanji conversion method of the present invention. 1: Input section, 2: Word search section, 3: Dictionary, 4: Temporary dictionary, 5: Composite word generation section, 6: Temporary dictionary registration section, 7: User selection section, 8: Output section. 1 person out

Claims (2)

[Claims] (1) It has a dictionary means for storing conversion information for converting input pronunciations into kanji or kanji-kana mixed sentences, and converts words not stored in the dictionary means into words stored in the dictionary means. In a kanji conversion method that generates a compound word, the dictionary means has a first dictionary means and a second dictionary means, and at least the generated compound word adopted as a conversion result is stored in the second dictionary. A kanji conversion method that is characterized by registering into means. (2) In the kanji conversion method according to claim 1, when registering in the second dictionary means, a character string portion of a predetermined type in a compound word is written in a form indicating an arbitrary character string of the type. A kanji conversion method characterized by registration.