JP2695772B2 - Kana-Kanji conversion device - Google Patents

Description

The present invention relates to a kana-kanji conversion apparatus for converting kana input into a kana-kana mixed sentence, and particularly to a kana-kanji input sentence that is not separated in phrases. The present invention relates to a kana-kanji conversion device that performs conversion following input without special conversion start instruction means such as a key. [Prior Art] Conventionally, there have been known several methods of converting a sentence of kana input (so-called solid input) which is not written in units of phrases into a sentence mixed with kanji kana. There is a method described in Kana-Kanji conversion device in Sho 60-189565. The method shown here is to convert kana input accurately to sentence mixed with kanji kana, and if there is ambiguity in the result of conversion, efficiently extract and hold multiple candidates for it, and Provides a means to quickly and easily select the correct conversion result from In addition, the method described in the Nikkei Computer November 25, 1985 issue of "Full Text Kana-Kanji Conversion Method Technology" uses a method of inputting kana character strings Independent dictionary matching with substrings is performed, and simple phrase matching is performed. As a result, when a certain amount (up to 4 phrases) of kana character strings is recognized, the regular kana character string is automatically recognized. This is a method of starting kana-kanji conversion. This is a method in which conversion is performed without the operator pressing the conversion key and instructing to start conversion, thereby facilitating recognition of the operator. Further, in Japanese Patent Application Laid-Open No. 60-22226, "Kana-Kanji conversion device" performs a dictionary search with the longest match each time a character is input,
A method is described in which kana-kanji conversion is immediately performed at high speed. [Problems to be Solved by the Invention] Of the three prior arts described above, the first one needs to first input a kana sentence to some extent, then press a conversion key to instruct the start of kana-kanji conversion, It takes time from pressing the conversion key until the conversion result is obtained, and even if it is not a phrase-based conversion method, the operator must press the conversion key with awareness of the break of the input sentence, which is troublesome There was a problem. The second conventional example is to solve this problem. However, since the conversion is started after the number of input phrases has reached 4 or more, the problem of the time required to obtain the conversion result has not been completely solved, and the extraction of four phrases in the preprocessing is complete. However, it is necessary to make corrections in the subsequent conversion process, and there is a problem that there is an overlap in the processing. Further, all the independent words are read from the dictionary before the conversion processing. However, many of them are unnecessary in the subsequent processing, so that this part is wasteful in the processing. In the last example, the conversion process can be performed at high speed, but dictionary matching is performed exclusively with the longest match only.
There is a problem in the conversion accuracy as compared with the first conventional example. The present invention improves the conversion accuracy and the processing time, which have been difficult to solve in the kana-kanji conversion according to the prior art, and provides a high-speed and high-precision conversion processing means. It is an object of the present invention to provide a kana conversion device capable of easily extracting another candidate even when making a correction. [Means for Solving the Problems] In order to achieve the above object, the following means are used. (1) A keyboard for inputting characters including kana, a memory for storing the input character string, a dictionary storing information such as word notation and part of speech as a means for driving the reading of the word, and the input character string A dictionary search means for extracting a partial character string from the memory to search for a word having a corresponding reading in the dictionary, and a means for performing a kana-kanji conversion including a connection test between preceding and succeeding words based on the searched word; Means for storing a plurality of kana-kanji conversion candidates generated by the conversion process, means for displaying the conversion candidates in order of likelihood, and means for selecting a correct word from the displayed candidates, and the dictionary search means Directly reads the character string in the input character string memory, performs character input processing when the character in the memory is insufficient, and sequentially and alternately processes the input of the character string and the dictionary search. Add to input And providing means for performing kana-kanji conversion. (2) A means for providing a backward key (backspace key) for correcting the immediately preceding input error and for deleting the last character in the input character string memory which has already been entered in response to the input of the backward key. Means for deleting the conversion candidate containing the input character correspondingly deleted from the conversion storage means, and adding a new conversion candidate from the end of each conversion candidate or the head position of the conversion candidate sequence remaining in the conversion candidate storage means. Means for starting creation of the conversion candidate is provided. (3) Further, at the start of the process in which a plurality of candidates are generated as described above, the means for storing and managing the start of the conversion process for each candidate, and the executable process stored in the storage management means are sequentially described. Means for taking out and executing the processing, and registering in the storage management means that this processing has entered the input waiting state when the processing being executed is waiting for input of a new input kana character string during dictionary search. In addition, there is provided means for retrieving another executable process in the execution waiting state and resuming the execution. [Operation] According to the first means (1), it is possible to perform a conversion process in parallel with the input of the kana character. Therefore,
As in the first prior art (Japanese Patent Application Laid-Open No. 60-189565), there is no need to take a procedure such as executing a conversion by inputting a character string to some extent and then pressing a conversion key, and without making the operator wait. Can perform processing. In addition, this means can be incorporated as a part of the above-mentioned conventional example, whereby a high-accuracy and high-speed kana-kanji conversion apparatus can be realized. The second means (2) describes internal processing means for the function of the reverse key, and is realized by linking the respective means described herein. In addition, by starting re-conversion from a possible position, it is possible to cope with conversion for a new input character after deletion. The third means (3) is for increasing the processing efficiency and speeding up by expanding the range of parallel processing and performing more thorough parallel processing. By means (1) and (3), the conversion processing can be performed more quickly and immediately than in the second conventional example. Examples Hereinafter, the present invention will be described in detail with reference to examples. FIG. 1 shows an embodiment of a kana-kanji conversion device according to the present invention. 1 is an input keyboard, 2 is an input character string memory for storing input character strings, 3 is a conversion dictionary, 4 is a dictionary search unit, and 5 is a kana-kanji conversion based on words obtained from the dictionary search unit. A conversion unit, 6 is a storage unit for conversion candidates, 7 is a display selection control unit for controlling display and selection of conversion candidates,
Reference numeral 8 denotes a display unit that displays conversion candidates on a screen in a viewable manner. FIG. 2 shows a kana-kanji conversion unit 5 and a conversion candidate storage unit in FIG.
6. The processing in the display selection control unit 7 is collectively a flowchart. When the conversion process is started, first, in step 201, a pointer i indicating the phrase conversion start position in the input character string memory 2 is entered.
Is set to the initial value 0, and positioned at the beginning of the input character string (in a state where no characters have been input yet). Then step 202
Check if the position is a break between clauses, that is, the end of a clause. Since the beginning of the input character string can be regarded as a phrase end, in this case, the process proceeds to phrase conversion in the next step 203. In the phrase conversion, as can be seen from the configuration diagram of FIG. 1, characters such as kana are input from a keyboard and are sequentially converted into morphemes constituting the phrase. Here, a phrase is centered on one independent word, preceded by one or more optional prefixes, followed by one or more optional suffixes, and followed by one or more optional prefixes. It refers to a form with a series of attached words. Compound words are divided into blocks containing one independent word that composes them, and each is regarded as a clause. For example, for "industrial application field", "industrial" (independent word 1, suffix 1, appendix 1), "use""field"
(Each independent word 1) is one phrase. As a method of kana-kanji conversion for each phrase, NHK Technical Research No. 25
The method shown in “Volume No. 5 (May 1973)“ Kana-Kanji Conversion by Computer ”” is well known, and the method can be basically applied here. Here, when one complete phrase is extracted, the likelihood of the probability of the conversion candidate is determined in step 204, and it should be stored in the conversion candidate storage unit 6 or discarded (called "branch"). Decide. The likelihood of certainty is the IPSJ 1983
19th Annual Meeting 5E-4 "Kana-Kanji conversion system for solid written sentences", 1981 IPSJ Computational Linguistics Research Document 25-6, "Table-based phrase structure analysis algorithms and their efficiency." And so on, the input character string is determined so that the likelihood is higher when the input character string is decomposed into a sequence of longer phrases, or in other words, a sequence of fewer phrases. However, adnominals such as "this" and "that" and formal nouns such as "koto" and "mono" are often used in conjunction with other clauses, so independent clauses such as nouns and verbs When counting the number of clauses without considering it, it is set to a value smaller than 1. As described above, the number of phrases is obtained by weighting in consideration of the part of speech and the appearance frequency, and it is assumed that the smaller the number is, the higher the likelihood is. Specifically, weights for nouns, verbs, adjectives, adjective verbs, etc., and weights for formal nouns, auxiliary verbs, adverbs, etc.
0.1, prefixes and suffixes are treated as semi-independent words, and a weight of 0.5 is given. Pruning calculates the likelihood from the beginning of the character string to the end of the phrase that is currently the subject of the determination, and determines it as the likelihood at the character position at the end of the phrase. If the degree is adequate, the value is compared with the value, and if the value exceeds a certain allowable value, pruning is executed. As the allowable value, for example, the minimum value + 1 is used for the number of weighted phrases at the same character position. Each clause that passed the pruning in step 204 is referred to as step 2.
At 05, it is stored in the conversion candidate storage unit 6. Also, step
At 206, the phrase sequence of the candidate with the highest likelihood among the phrases obtained so far is displayed on the screen of the display unit 8 (actually, the most likely likelihood is Higher candidates are added and displayed). In step 207, 1 is added to i, and the pointer is advanced to the next input character position. At step 208, it is determined whether or not there is any input character at the character position of the pointer i. If there is, the process returns to step 202. The input of characters is performed in parallel during the phrase conversion in step 203. If the input has been made continuously, the end of the input cannot be reached in step 208, so the loop of FIG. 2 is repeatedly processed. Even if the input of characters has been completed, the loop is repeated if unprocessed characters remain in the input character string. The above loop is repeated, all the input characters are fetched, and when the conversion processing for them is completed, the step
Proceeding to 209, selection input is received from the operator, and the intention of the operator is sequentially selected from among the candidates of each part to determine the result. Although the details of the selection operation and processing are not described here, for example, it can be carried out by the method shown in JP-A-60-189565. FIG. 3 shows a specific example of data formed in the conversion candidate storage unit 6 in accordance with the processing described above. The input character string taken as an example is "I'm going to do this", which is shown at the top of the figure. The lower part of the figure shows data of a plurality of conversion candidates created as a result of the conversion processing. In order to prevent duplication of the conversion candidate data, a network-like data structure is adopted to save the storage capacity. FIG. 4 shows a processing flow of the phrase conversion in step 203 of FIG. It consists of prefix processing (step 311), independent word processing (step 312), suffix processing (step 313), and attached word processing (step 314). As is clear from the figure, only the independent word processing is essential, and the other processing can be omitted. This is a schematized version of the definition of clauses described above, and is actually a prefix, suffix,
For each processing of the attached word, the processing is executed for all combinations of the case where the processing is performed and the case where the processing is not performed, and an attempt is made to extract all possible phrase conversion candidates. In addition, not only one attached word but also a plurality of attached words may be present, but all of them are tried. In these respects, FIG. 4 is different from the usual expression of the flow chart and is for the purpose of explanation. In each process of FIG. 4, a word having a pronunciation starting from the kana character indicated by the pointer i on the input character string memory 2 is extracted while collating with the dictionary, and sufficient characters are arranged on the input character string memory 2. If not, wait for an input character from the keyboard 1. In addition, a word that matches with the dictionary is extracted, and a connection check with the preceding word is performed. FIG. 5 shows a flow of the independent word processing. Prefix processing, suffix processing, and auxiliary word processing have the same basic structure as that of the independent word processing. In step 321, a request is first sent to the dictionary search unit 4, and words matching the partial character strings in the input character string memory 2 are read from the dictionary 3. At this time, all words having the same character position at the beginning (the value of the pointer i is the same) on the input character string memory 2 are extracted. In the example of FIG. 2, “nest”, “vinegar”, “number”, “suck”, “mathematics”, etc. are extracted as conversion candidates from the top. Next, in step 322, the conversion candidates extracted are extracted one by one, and in step 323, a connection check with the preceding morpheme (word) is performed. In the above example, since the extraction start point is at the beginning of the input character string, that is, at the beginning of the phrase, all the independent words satisfy the connection condition and the "nest"
"Vinegar", "number", "suck", "mathematics", etc. passed the check in step 324 and were recognized as part of the phrase candidates. Then, in step 325, after the existing part (in this case, the head)
Connected to In the example of the general position, there is a portion where the particle “ga” (corresponding to an adjunct) is cut out in FIG. 3, but the particle “ga” is a case particle or a connecting particle, and each particle is a noun. Since it is possible to connect with the nomenclature such as verbs and verbs, as shown in the figure, the case particle "ga" is cut out for "number", and the connecting particle "ga" is cut out for "suck" as shown in the figure. You. When the connection check is completed for all the conversion candidates extracted in this way (step 326), the independent word processing ends. Prefix processing, suffix processing, and adjunct processing are almost the same. FIG. 6 shows the processing flow of the dictionary search unit 4 in step 321 in FIG. In step 401, a pointer k representing the end of the reading of a word to be dictionary-searched in the input character string memory 2 is initialized to the value of the leading end position pointer i of the word. Next, in step 402, it is checked whether or not the input character has already been registered at the position of the pointer k on the input character string memory 2, and if there is, the process proceeds downward. If there is no input character, it waits for a character to be input from the keyboard (step 403). When a character is input, it is read and registered at the position of the pointer k in the input character string memory 2 (step 404). In step 405, the dictionary is searched by reading the character string in the section between i and k. As a result of the search, when a matching word is obtained (step 406), it is registered in a dictionary search word buffer (step 407). Here, the dictionary search word buffer is a memory for temporary storage provided in the dictionary search unit 4.
Next, at step 408, the pointer k is advanced by one, and the process returns to the beginning of the loop. If no matching word is found in the dictionary in step 406, the processing exits from the loop.
The words registered in the dictionary search section buffer are passed to step 322 in FIG. If no matching word is found in the dictionary in step 406, the process is terminated.However, in this process flow, the process is performed from the short reading to the long reading, so that the matching reading is not performed. Even if the word doesn't happen,
If the reading length is further increased, a matching word may be found. For example, when the character string “Sugaraku” is in the input character string memory 2, the words for “Reading” include “Number” and “Suck”, but there is no word for “Sugar”. However, if you extend the reading further to "sugar," there is the word "mathematics." To cope with such a situation, the structure of the traction unit of the dictionary may be a tree structure of the reading order as shown in FIG. 7, for example. Speaking of the above example, for reading "su", "nest"
It can be seen that “vinegar” and “su” have words such as “number” and “suck”. Although there is no corresponding word for "Suga", it is possible to easily determine that there is a matching word when the reading is further extended. In addition, when "ku" of "sugaraku" comes, it is immediately understood that there is no matching word before that. By utilizing this point, it is detected that there is no longer word when "mathematical" is searched, and the flow shown in FIG. Can also be changed. In the case where words having a plurality of lengths corresponding to a given character string are obtained as shown in FIG. 6, the words must be stored in a buffer (a dictionary search word buffer) of each of the words obtained as a result of the search. Instead, it is possible to take control such that each time a word is obtained, the word is passed to the next step (in this case, step 322 in FIG. 5). In this way, other processes such as connection check can be executed in parallel without waiting until all words are obtained from the dictionary, so that the processing efficiency can be improved. By the way, independence words often have longer reading lengths than prefixes, suffixes, or adjunct words. In this case, it is often useless to perform processing such as a connection check for all of the short to long readings. Practically, even if it is limited to words obtained by the longest match or the next longest match (the word having the same length after reading the longest match word), there is almost no problem in the accuracy of kana-kanji conversion. Therefore, in this case, it is necessary to increase the reading length one by one until the longest match is obtained, and then search for the next longest match after the longest match is obtained. I can do it. As described above, the example in which the character is input without error has been described. However, when a human is inputting the character from the keyboard, sometimes an input error is made. However, input mistakes are often noticed on the spot. To correct this, the keyboard is usually provided with a back key for canceling the immediately preceding input character. Next, the operation related to the backward key will be described. FIG. 8 shows a flowchart of the reverse key processing. When the backward key is pressed, first, at step 341, one character at the end of the input character string memory 2 is deleted. Then step
At 342, a dictionary access cancel including the character being executed by the dictionary search unit 4 is canceled. In step 343, the conversion candidate of the morpheme (word) including the character deleted in the conversion candidate storage unit 6 is deleted from the network. Finally (step 344), the conversion process is attempted again from the end of each morpheme (word) on the network and the beginning of the conversion character string. In this case, it is not necessary to perform the processing again for the morphemes (words) already cut out, and it is sufficient to perform the conversion processing, that is, the dictionary search for the character string including the newly input character from the part of the deleted character. . If the maximum length of a word in the dictionary is longer than the length of the input character string, there is no need to perform re-conversion processing from the end of all morphemes (words) on the network and the beginning of the conversion candidate string. The re-conversion processing may be performed in a range from the last end of the word to the point where the word in the dictionary is read back by the maximum length. This is the end of the description of the first embodiment. By the way, as is apparent from FIG. 3, when an input character string to be converted is generally given, there are a plurality of kana-kanji conversion candidates corresponding to the input character string. In addition to mere homonyms, there may be a plurality of candidates for how to divide an input character string into words or for its part of speech. At this time, the first
The processing of each part in FIG. 1 shown in the embodiment of FIG.
There is no problem if the processing required for each character input is performed and the processing is completed before the next input is performed, but in reality, the processing of each part in FIG. 1 requires a finite execution time. This may cause a situation in which the processing speed cannot keep up with the input of characters from the keyboard. This occurs, for example, when a skilled operator inputs characters at a very high speed, or when a large number of conversion candidates that can exist in parallel occur. A second embodiment capable of efficiently performing kana-kanji conversion even in such a situation will be described below. FIG. 9 shows the configuration of this embodiment. In FIG. 9, 1
8 are the same as those in FIG. Is a parallel processing control unit for switching and executing the processing of a plurality of candidates existing in parallel one after another. The operation of the kana-kanji conversion device according to the present embodiment will be described using the specific example shown in FIG. This example shows a state in which "sugakuka" has been input in the process of finally obtaining the result shown in FIG. In the part of “Suugaraku”, conversion candidates are clarified, and “K” and the subsequent processing have been started. As for "ka", the auxiliary particle "ka"
Alternatively, there are suffixes such as "family" and "ka", which are already registered in the network. Further, since the phrase can end at this position, the processing of the next phrase is started from the rear end. It also indicates that the processing of a phrase consisting of "?" And an input character string following it has been started. While the kana-kanji conversion is being executed in the embodiment of FIG. 1, it is assumed that a situation as shown in FIG. 10 is reached when a keyboard input wait state is entered in steps 402 and 403 in FIG. At this time, the auxiliary particle "ka" and the suffix "ka"
There is no problem when processing of a clause consisting of "ka" and the input character string following "ka" after "ka" is cut out, but this is the opposite before "ka" or "family""ka" is cut out In the middle of the phrase processing for the input character string following "ka", it enters a state of waiting for input, and is kept waiting even though the processing of "ka" or "family""ka" can be executed In some cases. In such a case, if it is detected that the processing of one clause has entered the state of waiting for input, and another executable processing is extracted and executed using the idle time, either of the processings can be performed. Even if the processing is performed first, it is possible to execute all the processing that can be executed at that time. The following are candidates for such parallel processing utilizing the state of waiting for keyboard input. (1) Plural phrase processing on the input character string memory 2 with different phrase start points, and subsequent processing. (2) A plurality of processes having different connection relations before and after, such as a prefix, an independent word, a suffix, and an attached word. In addition, multiple processes based on polysemy, such as part of speech and inflection. (3) A plurality of processes for words having the same starting point but different ending points, that is, words having different lengths. For each of these, when the ambiguity occurs, the start of execution of each process is notified to the parallel processing control unit 9 and registered. When each process waits for a keyboard input while accessing the dictionary, the parallel processing control unit 9 determines whether the registered process is a registered process. Select the executable from among them and start it. The time when polysemy occurs is, for example, when the phrase conversion is started while moving the pointer i in FIG. 2 to check for the presence or absence of a clause end. In FIG. 4, prefix processing, suffix processing, For example, there is a branch portion regarding whether or not the process is executed. [Effects of the Invention] As described above, according to the present invention, in a device for inputting a kana character string that is not divided and written in units of phrases and converting it to a kana-kanji character, without using a special conversion key, Kanji conversion can be performed, and the conversion is performed following the input, so that it is easy to confirm the conversion result and a kana-kanji conversion device with good operability can be realized. Further, by incorporating the present invention as a part of the conventional example (Japanese Patent Laid-Open No. 60-189565), it is possible to realize a high-accuracy and high-speed kana-kanji conversion device. It should be noted that the present invention has a configuration in which kana character strings that are sequentially input are sequentially converted, and is applicable to a device in which kana character strings are sequentially input such as an interactive voice input device. Needless to say, it is effective.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a diagram showing a first embodiment of the present invention, FIG. 2 is an overall flow of a kana-kanji conversion process, and FIG. 3 is formed in a conversion candidate storage section 6 by this process. It is a specific example of data to be performed.
FIG. 4 is a processing flow of phrase conversion, FIG. 5 is a processing flow of independent word processing, and FIG. 6 is a processing flow of the dictionary search unit 4. FIG. 7 shows the configuration of a dictionary pulling unit used in the processing flow of FIG. FIG. 8 is a flowchart of the reverse key processing. FIG. 9 shows the configuration of the second embodiment.
The figure shows a specific example for explaining the operation. 1 ... keyboard, 2 ... input character string memory, 3 ... dictionary, 4
... Dictionary search unit, 5 ... Kana-Kanji conversion unit, 6 ... Storage unit of conversion candidates, 7 ... Display selection control unit, 8 ... Display unit, 9 ... Parallel processing control unit.

Claims (1)

(57) [Claims] A keyboard for inputting characters including kana, an input character string memory for storing the input character strings, a dictionary for storing information including at least one of a word notation and a part of speech, and a partial character string from the input character string memory. A dictionary search means for extracting a word having a corresponding reading in the dictionary and performing a kana-kanji conversion process including performing a connection test with a preceding word based on the searched word; Means for generating kanji conversion candidates, means for storing a plurality of kana-kanji conversion candidates generated by the kana-kanji conversion process, means for displaying the plurality of kana-kanji conversion candidates in the order of likelihood, and displayed kana Means for selecting a correct word from kanji conversion candidates, and the dictionary search means sequentially searches the dictionary in parallel with input of characters from the keyboard. What is claimed is: 1. A kana-kanji conversion device, comprising: a parallel processing means for selecting an executable process while waiting for an input from the keyboard and starting the process as parallel column processing. 2. 2. The kana-kanji conversion device according to claim 1, wherein the execution of the dictionary search is started when an ambiguous character occurs in the input character string. 3. 3. The kana-kanji conversion device according to claim 1, wherein the dictionary has an index portion having a tree structure in reading order. 4. In claim 1 or 2, the means for performing the kana-kanji conversion process determines the likelihood such that the likelihood is higher when the input character string is decomposed into a longer phrase string, A kana-kanji conversion device, wherein a plurality of kana-kanji conversion candidates are generated. 5. 3. The kana-kanji conversion device according to claim 1, wherein the means for storing the plurality of kana-kanji conversion candidates has a network-like data structure. 6. A keyboard for inputting characters including kana, an input character string memory for storing the input character strings, a dictionary for storing information including at least one of a word notation and a part of speech, and a partial character string from the input character string memory. A dictionary search means for extracting a word having a corresponding reading in the dictionary and performing a kana-kanji conversion process including performing a connection test with a preceding word based on the searched word; Means for generating kanji conversion candidates, means for storing a plurality of kana-kanji conversion candidates generated by the kana-kanji conversion process, means for displaying the plurality of kana-kanji conversion candidates in the order of likelihood, and displayed kana A means for selecting a correct word from the kanji conversion candidates; and a parallel processing means for selecting a process that can be executed while waiting for input from the keyboard and starting the process as parallel column processing. A kana conversion device configured to sequentially perform a search of the dictionary by the dictionary search means in parallel with input of a character from a keyboard, wherein a reverse operation is performed to cancel a character input immediately before on a part of the keyboard. Means for deleting the last character in the input character string memory which has already been input in response to the input of the backward key, and a plurality of kana-kanji conversion candidates for freezing the deleted character. Means for deleting from the storage means for storing the conversion candidates; and at least one of the end of each kana-kanji conversion candidate and the beginning of each kana-kanji conversion candidate remaining in the storage means for storing the plurality of kana-kanji conversion candidates. Means for performing a kana-kanji conversion process again, and a kana-kanji conversion device.