JPS62271173A - Kana/kanji conversion system - Google Patents

Abstract

PURPOSE:To improve the processing speed without changing the size of a dictionary by registering a word having a long reading at first in the dictionary if reverse sequential retrieval is more difficult than the forward retrieval or impossible in the word data retrieval in the dictionary. CONSTITUTION:In registering the representation of a word, the words are registered in the order of Japanese syllabary of the read of the head of each word and a word having a longer reading is registered earlier in the words having the same head reading. That is, when the lengths of the reading of optional words, A,B in the dictionary are a,b, respectively in case of the relation of a< and the reading of the head character of the word B is coincident with the reading of the word A, the word B is registered earlier without fail in the forward retrieval. It is assumed that the sentence KANA(Japanese syllabary)/KANJI(Chinese character) conversion is processed while being regarded as (reading to be converted)=(reading registered in dictionary)+(declension word end, reading of adjunct), where the relations of (reading registered in dictionary)>= 1 and (reading of declension word end and adjunct) >=0 exist.

Description

[Detailed Description of the Invention] 3. Detailed Description of the Invention Technical Field The present invention relates to a kana-kanji conversion method, and in particular, to improve the conversion speed of both long and short words in a word processor, etc. This paper relates to a kana-kanji conversion method that allows for

Prior Art In conventional dictionaries, words are generally registered in order of reading code. In other words, in a dictionary arranged in iweo order, words with short reading lengths are registered first.

Therefore, for example, if we compare "core" and "follower",
``Pashin'' is registered first.If you want to convert the kana ``Shintopa'' into kanji, the dictionary search will be ``Shinto'',
It is necessary to do three things: ``shin'' and ``shi''. Between the expressions ``believer'' and ``core,'' the former has a longer pronunciation of the word registered in the dictionary, so it has a higher probability of being correct, and is the first to be output. However, if the dictionary is searched in the order of the longest reading length, the processing speed will decrease because the search will be in the backward direction.

Furthermore, if the search is performed in order of shortest length, notations cannot be output until the dictionary search is completely completed, and the required candidate storage area becomes larger, which is not preferable. Also.

If a dictionary is allowed to be sequentially searched in both forward and reverse directions, there is a problem in that the dictionary size increases.

FIG. 1(a) is a diagram showing the conventional dictionary registration order. The reading length of arbitrary words A and H in the dictionary is a.

When b, if a < b and the pronunciation of the first character of B matches the pronunciation of A, then B is always registered later in the forward search. In other words, Figure 1 (
As shown in a), words with long reading lengths such as ``table Japan'', ``baggage'', ``'Japan'', and ``hon'' will be registered later, and words with short reading lengths such as ``go'' and ``book'' will be registered later. Words such as ``tail'', ``hand'', ``2'', and ``hang'' are registered first.

Purpose The purpose of the present invention is to provide a kana-kanji conversion method that can improve the above conventional problems, improve processing speed without changing the size of the dictionary, and simplify the dictionary search algorithm and candidate word accumulation algorithm. Our goal is to provide the following.

Configuration In order to achieve the above object, the kana-kanji conversion method of the present invention is stored in the main memory or auxiliary memory, and when searching for word data, searching in the backward direction is more difficult or difficult than in the forward direction. In a kana-kanji conversion system that has an impossible dictionary and a control means that performs kana-kanji conversion processing using the dictionary, between words whose first letters in the dictionary have the same reading, words with a long reading length are The feature is that the words are registered first and the words in the dictionary are searched sequentially starting with the words with the longest pronunciation length.

Hereinafter, the configuration of the present invention will be explained in detail using examples.

FIGS. 1(b) and 1(c) are diagrams showing the dictionary registration order according to the first and second embodiments of the present invention. In the first embodiment, when registering the notation of a word, it is registered in the order of the first pronunciation of the word, and among words with the same first pronunciation, the one with the longer pronunciation is registered first. Note that the notation in parentheses is a conjugated ending and is not registered in the dictionary. In Embodiment 2, when registering the spelling of a word, the first pronunciation of the word is in the reverse order of the pronunciation of the word, ``n'', ``wa'', ``mouth'', ``shi'', ゛ru'', etc. Register in this order. And among those with the same initial pronunciation,
This is the same as the first embodiment in that the longer reading length is registered first. That is, in the present invention, when the pronunciation lengths of arbitrary words A and B in the dictionary are a and b, a (b
And if the reading of the first character of B matches the reading of A, then B is always registered first in the forward search.

In this example, for simplicity, it is assumed that bunsetsu kana-kanji conversion is processed as follows: (reading to be converted) = (reading registered in dictionary) + (reading of conjugated endings and attached words). Here, (dictionary reading) ≧1. (Inflection endings, pronunciations of attached words) ≧0.

FIG. 2 is a block diagram of a kana-kanji conversion device showing an embodiment of the present invention. In Figure 1, 1 is an operating section such as a keyboard for inputting the ``kana'' to be converted, 2 is a display section such as a CRT display for displaying and outputting the converted ``kanji'', and 3 is a dictionary search unit. 4 is a dictionary search unit that searches the dictionary by referring to the index; 5 is a sequential search unit that reads word data directly from the dictionary; and 6 is a control unit that controls word endings and adjuncts. Creation section to create.

7 is an index search unit for searching an index, 8 is a dictionary, and 9 is an index.

First, when data to be converted is input from the operation section l,
The control unit 3 converts the readings to be converted into dictionary readings, conjugated endings,
It is divided into the pronunciations of the attached words, and the dictionary search unit 4 obtains the notation of the word corresponding to the dictionary pronunciation, the inflected ending, and the notation of the pronunciation of the attached word. The dictionary search unit 4 uses the index search unit 7 to obtain a dictionary pointer, and then sequentially instructs the search unit 5 to search the dictionary. The sequential search unit 5 searches the dictionary 8 in the order in which words are registered, and obtains word data for dictionary readings. The index search unit 7 searches the dictionary index and obtains the position (dictionary pointer) of the block containing the desired word data. Also,
The ending/adjunct word creation unit 6 creates inflected endings and adjunct words for the word obtained by dictionary search.

First, a case where the present invention is applied to phrase kana-kanji conversion will be explained.

Assume now that the contents shown in FIG. 1(b) are registered as the dictionary 8 in FIG. 2. The operation when converting kana to kanji will be explained using ``Paomotenihon'' as an example of the reading to be converted.

(1) The control unit 3 searches for candidates with the dictionary pronunciation as "Omote nihon" and the pronunciations of endings and adjuncts as absent.

(2) Next, the dictionary search unit 4 searches for word data for "Omote Nihon". In this case, first, a dictionary pointer for "Omote Nihon" is obtained through an index search, and then word data for "Omote Nihon" is obtained through sequential searches. At this time, the dictionary pointer points to the first-order word table ``''.

(3) Obtain the candidate notation ``Omote Nippon'' for this ``Omote ni Hon''.

(4) Next, the control unit 3 reads the dictionary on the front page.
Search for candidates using the reading of the ending and attached word as ``n''.

(5) The dictionary search unit 4 searches for word data for "Omoteha". In that case, according to the registration order of the present invention, words with the reading ``Omote wa'' must be registered after words with the reading ``Omote nihon,'' so no index search is necessary. be. By sequential search, ``On the front''
Detects that there is no word data for. then,
The dictionary pointer does not change.

(6) In the same way as in (4) above, candidates are searched for using the dictionary reading as ``main'' and the readings of the endings and EX words as ``han'', but there is no word data for this either.

(7) Next, search for candidates using the dictionary reading as ``Omote'' and the readings of the endings and adjuncts as ``Nihon.''

(8) The dictionary search unit 4 obtains the word data of the ``table'' for the ``front''.The dictionary pointer is the primary word ''
It refers to thoughts.

(9) As a notation for attached words, get pa ni honpa.

(10) Candidate notation for “Omote Nihon” Obtain “Table 1”.

Hereafter, in the same way, candidates are searched for ``Omo 11 + #ShiNihon'°, 'O... + 11MoTeNiHonpa.

FIG. 8 is a comparison diagram showing the search procedure of the conventional dictionary and the dictionary of the present invention regarding dictionary search when converting "Omote Nihon".

In the present invention, when the first characters have the same pronunciation, the longer pronunciation of the word is registered first, so the dictionary pronunciation of the pronunciation to be converted is sequentially shortened, and Since the dictionary search is performed by sequentially lengthening the pronunciations, it is only necessary to search for words in the dictionary after the position where the word was previously searched, thus eliminating the need for an index search. In that case, the dictionary pointer points to the previously searched position and does not change, so the next search can be restarted from that position.

That is, as shown in FIG. 8, in the first search for ``Paradise Japan'', the conventional method and the present invention use the same procedure to obtain index search -> sequential search -> ``Table Japan''.

After the next search for ``,'' as mentioned above, in the present invention, it is possible to omit the index search when searching for nearby words during dictionary search, so sequential search → word Data can be obtained and speeding up can be achieved.

Next, regarding the case where the present invention is applied to complex clause conversion,
explain.

In compound clause conversion, candidates are created by combining multiple words, so it is necessary to accumulate word data obtained through dictionary searches.

In general, in compound clause conversion, candidate notations with a long reading length of the first word are output with priority. Therefore, when searching the dictionary, it is easier to search in the order of long reading length when the storage area is full.

Therefore, as in the case of bunsetsu conversion, the dictionary search algorithm searches for items with the longest reading length, and speeding up can be achieved by omitting index search.

Next, regarding the case where a buffer for reading 8 is provided,
explain.

If the dictionary exists in auxiliary storage, reading the dictionary will be done through a buffer. Therefore, the processing speed of successive searches for adjacent words is significantly improved by the present invention compared to the conventional method.

FIG. 3 is an operation sequence chart of the control section in FIG. 2. In FIG. 3, input data 11 is data input from the operation unit 1 and consists of a mode and reading. The output data 12 is data indicating a conversion result output from the control section 3 to the display section 2, and consists of a mode and a description. 1
3 is a candidate notation, which is sent from the ending/adjunct word creation unit 6.

14 is search information, dictionary search reading, dictionary pointer 5
and search results. 15 is word data, which is sent from the dictionary search section 4.

When the control unit 3 controls kana-kanji conversion, it first checks the mode of the input data 11, and when it is in the conversion mode, sets the dictionary search reading from the input data 11 and clears the dictionary pointer. When in the next candidate mode, it waits.

In addition, for the output data 12, set the output mode (determined mode), word ending, and candidate notation 1 from the adjunct word creation unit 6.
When 3 is sent, the notation of output data 12 is set.

Then, as a result of searching the dictionary using the search information, if the word is found, the word and m-word are created, the candidate notation output mode is set, and the notation is set.

In addition, if there is no word, check the dictionary search reading length,
When it is 1 or more, the reading is shortened. Further, when the reading length is 0, a no-candidate display mode is set.

FIG. 4 is an operation sequence chart of the dictionary search section in FIG. 2. The dictionary search section 4 exchanges search information 21 with the control section 3 and sends word data 22 that is the search result to the control section 3. First, when searching a dictionary, initialize the dictionary pointer (that is, check whether the pointer has been cleared). When cleared, search the # index. When the dictionary pointer is sent from the index search unit 7, it instructs the sequential search unit 5 to perform a sequential search. The results of the sequential search are placed on the word data 22, and the search results are set in the search information 21.

FIG. 5 is an operation sequence chart of the index search section in FIG. 2. The index search unit 7 uses a dictionary pointer 24 to receive and receive dictionary search readings 23 from the dictionary search unit 4.
to be sent back. Then, the index 9 is retrieved by comparing the index 25 consisting of an index reading and a dictionary pointer. First, after searching the dictionary index, the dictionary search reading 23 and the index reading of the index 25 are compared. If they match,
A dictionary pointer within the same index 25 is set and sent to the dictionary search section 4.

FIG. 6 is an operation sequence chart of the sequential search section in FIG. 2. The sequential search unit 5 receives and receives dictionary search pronunciations 26 from the dictionary search unit 4, exchanges dictionary pointers 27, and sends word data 28. When the dictionary search unit 4 instructs the dictionary to be searched sequentially, the receiver compares the dictionary search reading 26 and the dictionary registered reading which is the content 29 of the dictionary 8.

If there is a match, word data corresponding to the content 29 is set and sent to the dictionary search unit 4.

FIG. 7 is an operation sequence chart of the ending and adjunct word creation section in FIG. 2. In the ending and adjunct word creation section 6,
Between the control unit 3 and the word data 30 and the dictionary search reading 31
``Uke'' takes the reading ``32'', searches the word ending list, refers to the cent ending length 34, creates a candidate notation, and then creates a candidate notation. 233 is returned. In other words, the ending/adjunct word creation unit 6
is equipped with a table of word endings, and when word data 30 obtained by dictionary search is sent from the control section 3, the creation section 6 creates word endings and adjuncts based on this. First, part of speech information of word data 30 and dictionary search reading 31
, the 1-word ending table is searched with reference to reading 32, and the length of the conjugated ending is thereby obtained (in the case of a predicate). Furthermore, if the word is not a predicate, the ending length is set to O. And word data 30
Based on the word notation, part of speech information, and pronunciation 32, the attached word is cented. A candidate notation 33 is created from the set R-word, ending length 34, and word notation, and sent to the control unit 3.

In this way, the operations of each section shown in FIG. 1 are executed. In kana-kanji conversion, it is necessary to give priority to outputting words with long reading lengths, but in this example, words with long reading lengths are registered first in the dictionary, which simplifies the processing algorithm. It is possible to improve the conversion speed.

Effects As explained above, according to the present invention, when it is difficult or impossible to search for word data in a dictionary in the backward direction compared to the forward direction, the word data search in the dictionary is difficult or impossible. Since words are registered first, processing speed can be improved without changing the size of the dictionary. Further, the present invention provides the longest -
It is suitable for conversion algorithms based on teaching methods, and can simplify dictionary search algorithms and candidate word accumulation algorithms.

[Brief explanation of drawings]

FIG. 1 is a comparison diagram of dictionary registration order showing a first embodiment and a second embodiment of the present invention, FIG. 2 is a block diagram of a bunsetsu kana-kanji conversion system showing an embodiment of the present invention, and FIG. Figure 4 is an operation sequence chart of the control unit in Figure 2, Figure 4 is an operation sequence chart of the dictionary search unit in Figure 2, Figure 5 is an operation sequence chart of the index search unit in Figure 2, and Figure 6 is the operation sequence chart of the index search unit in Figure 2. 2 is an operation sequence chart of the sequential search section, FIG. 7 is an operation sequence chart of the ending/adjunct word creation section in FIG. 2, and FIG. 8 is a comparison diagram of the dictionary search operation of 'Omote Nihon' according to the present invention. 1: Operation section, 2: Display section, 3: Control section, 4: Dictionary search section, 5: Sequential search section, 6: Ending/adjunct word creation section, 7: Index search section, 8: Dictionary, 9 :index.

Claims (1)

[Claims] 1. A dictionary that is stored in a main memory or an auxiliary memory and in which searching in the backward direction is more difficult or impossible than in the forward direction when searching for word data, and the dictionary is used. In a kana-kanji conversion system that has a control means that performs kana-kanji conversion processing using A Kana-Kanji conversion method that searches words sequentially starting from the longest reading length. 2. The dictionary stores words in the order of the first letter reading Aiueo or vice versa, and among words with the same first letter reading, the word with the longest reading length is registered first. A kana-kanji conversion method according to claim 1, characterized in that: