JP2008181357A - Japanese processing apparatus, japanese processing method and japanese processing program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To present conversion candidates adapted to a user even for a word which has not been input in the past. <P>SOLUTION: A Kana-Kanji conversion-prediction engine (core part) 42 determinates the priority order of conversion logics on the basis of the priority of respective conversion logics in the combination of conversion logics applicable to a character string to be converted and generates a conversion candidate list for the character string to be converted by using a Kana-Kanji conversion-prediction dictionary 44 in accordance with the priority order. The conversion logic determines an attribute of a word to be converted from the character string to be converted. When a conversion result is defined, a Kana-Kanji conversion-prediction engine (user dependence part) 43 increases the priority of the conversion logic used for extracting the defined notation for the combination of the conversion logics applicable to the character string to be converted. Namely, the Kana-Kanji conversion-prediction engine 43 learns which attribute is to be preceded in the combination of attributes allowed to be applied to words in a syntax allowed to be adopted to the character string to be converted. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a Japanese language processing apparatus, a Japanese language processing method, and a Japanese language processing program for performing Japanese language conversion processing on an input character string.

  In a Japanese processing method that performs conversion processing on input character strings such as readings entered in hiragana or romaji, such as kana-kanji conversion and predictive conversion, how to make the conversion candidate presentation suitable for the user On the other hand, various efforts have been made conventionally.

  For example, Patent Literature 1 discloses conversion candidates based on character string attribute information (such as whether a word means a place, whether a word means time, etc.) for classifying the content that the input character string means. The method of narrowing down is described.

  Further, in Patent Document 2, learning is performed by selecting one of a plurality of learning methods according to reading of an input character string to be converted, expression of a conversion result, part of speech, and the like, and the order of conversion candidates is set. The method of changing is described.

Japanese Patent Laying-Open No. 2005-107751 (paragraphs 0033-0035, 0045) JP-A-9-062668 (paragraphs 0049-0052)

  User adaptation in the conventional Japanese processing method is user adaptation at the word level that the user has input in the past, and there is a problem that it is not adapted to words that have not been input in the past. Here, the user application in the Japanese language processing method means changing the presentation contents of the conversion candidate according to the user input history (including the conversion result). For example, as a user adaptation method in the order of candidates, there is generally known a method of learning a notation that has been determined for the reading of a word input by the user and increasing the candidate order of the notation. However, the order of candidates is not changed for words that are performed only for words that have not been input in the past.

  The method described in Patent Document 1 is intended to support input in consideration of the situation where the user is placed, and is intended to improve user adaptability for words that have not been input in the past. is not. For example, if the method described in Patent Document 1 is applied, the conversion result can be narrowed down to words belonging to the attribute by inputting the character string attribute information even if the word has not been input in the past. . However, the method described in Patent Document 1 changes conversion candidates based on character string attribute information input at that time, and does not improve user adaptability based on user input history. Absent.

  Note that Patent Document 1 also mentions a method of automatically acquiring character string attribute information based on input character string information, but predicts that a word related to a place will be input next to “Go”. It only shows an example of “Yes”, and it does not disclose how to realize the prediction.

  In addition, the method described in Patent Document 2 stores a learning method corresponding to a specific case according to the usage status of the user, and switches the learning method for each specific case. The fact that the learning result by the switched learning method is reflected at the word level is the same as the conventional method.

  Accordingly, an object of the present invention is to provide a Japanese language processing apparatus, a Japanese language processing method, and a Japanese language processing program that can be adapted to a user even for words that have not been input in the past. Specifically, a Japanese language processing device, a Japanese language processing method, and a Japanese language processing capable of changing the conversion candidate presentation order to an order suitable for the user even for words that have not been input in the past The purpose is to provide a program.

  A Japanese language processing apparatus according to the present invention is a Japanese language processing apparatus that performs Japanese conversion on an input conversion target character string, and determines the attribute of the word that the conversion target character string can take and the conversion target character string. Japanese conversion means (for example, kana-kanji conversion-prediction device 4) that changes the priority of conversion logic that determines what attribute the character string to be converted is converted as a word based on the conversion result It is characterized by having.

  In addition, when the conversion result is confirmed for the conversion target character string, the Japanese conversion means is used to extract the notation that is the conversion result for the combination of conversion logics applicable to the conversion target character string. When a conversion logic learning means (for example, kana-kanji conversion-prediction engine (user-dependent unit) 43) for increasing the priority of the conversion logic and a conversion target character string are input, conversion applicable to the conversion target character string Conversion processing means (for example, kana-kanji conversion-prediction engine (core section) 42) that generates a conversion candidate list for the conversion target character string according to the priority of each conversion logic in the combination of logics may be included.

  Also, the conversion logic learning means, for a combination of two or more conversion logics that can be applied to the same character string, is a logic priority table for describing the priority of each conversion logic in the combination (for example, kana-kanji conversion − When a conversion result is determined for the conversion target character string, the conversion logic learning unit determines a combination of conversion logics applicable to the conversion target character string. Update the conversion logic used to extract the conversion result notation in the defined logic priority table to increase the priority of the conversion logic, and the conversion processing means will change the priority of each conversion logic shown in the logic priority table. Thus, a conversion candidate list for the conversion target character string may be generated.

  Further, the Japanese conversion means, as a kana-kanji conversion dictionary, stores a kana-kanji conversion dictionary storage means for storing a reading, each notation of the word having the reading, and the attribute of the word indicated by the notation ( Conversion processing means applies the conversion logic to the conversion target character string in descending order of priority of the conversion logic, and is specified by the applied conversion logic from the kana-kanji conversion dictionary. A conversion candidate list may be generated by extracting notations with matching word readings and attributes as conversion candidates.

  Further, the conversion logic learning means, for a reading registered in the kana-kanji conversion dictionary, a word priority table (for example, kana-kanji conversion-prediction) for describing the priority of each word indicated by the notation having the reading. When the conversion result is determined for the conversion target character string, the conversion logic learning means determines the conversion result in the word priority table determined for the word reading of the conversion result. When the conversion target character string is input, the conversion processing unit updates the priority of each conversion logic indicated in the logic priority table and the word priority table. A conversion candidate list for the conversion target character string may be generated based on the priority of each word indicated by.

  Further, the Japanese processing method according to the present invention is a Japanese processing method in which the Japanese conversion means performs Japanese conversion on the input conversion target character string, and a word attribute that the conversion target character string can take. Further, based on the conversion result determined for the conversion target character string, the priority of the conversion logic that determines what attribute the conversion target character string is converted to as a word is changed.

  In addition, a Japanese language processing program according to the present invention is a Japanese language processing program for performing Japanese language conversion on an input conversion target character string. The process of changing the priority of the conversion logic that determines what attribute the conversion target character string is converted to as a word based on the conversion result determined for the attribute and the conversion target character string. Features.

According to the present invention, the conversion order of priority of the conversion logic is based on the attribute of the word taken by the conversion target character string input by the user in the past and the conversion result determined for the conversion target character string. Therefore, even for words that have not been input in the past, the conversion candidate presentation order can be changed to an order suitable for the user.
.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing a configuration example of a Japanese language processing apparatus according to the first embodiment of the present invention. 1 includes an input device 1 such as a keyboard, an output device 2 such as a display device, a control unit 3 that controls input / output information, and kana-kanji conversion and prediction for an input character string. A kana-kanji conversion / prediction device 4 that performs conversion;

  The input device 1 inputs information indicating the contents of the user operation (specifically, a character string to be converted or information indicating a confirmation instruction for a conversion candidate) to the control unit 3 as input information. The output device 2 outputs the output information from the control unit 3 (specifically, an input character string, a list of conversion candidates, and information indicating the conversion result) so that it can be displayed to the user.

  The control unit 3 is a controller responsible for the interface with the kana-kanji conversion-prediction device 4, and the input information input from the input device 1 is a character string to be converted (for example, an unconfirmed reading input in hiragana or roman characters). In the case of the conversion target character string, various processes for causing the kana-kanji conversion-prediction device 4 to perform kana-kanji conversion and prediction conversion are executed. Here, kana-kanji conversion refers to conversion of a conversion target character string into a kana-kanji mixed expression using the conversion target character string as a reading. Predictive conversion refers to conversion of a conversion target character string into a kana-kanji mixed expression that uses the conversion target character string as part of the first half of reading.

  Specifically, the control unit 3 is realized by an information processing device that operates in accordance with a program such as a CPU in which a higher-level application of the kana-kanji conversion / prediction device 4 is read. For example, the control unit 3 sets a conversion target character string, acquires information on conversion candidates (conversion candidate list), or relates to a conversion result based on a confirmation instruction for the conversion candidates. Notify information. Further, output information is generated based on the information regarding the conversion candidate and the information indicating the confirmation instruction, and the output device 2 displays the information. After the conversion target character string is set, the kana-kanji conversion-prediction device 4 directly inputs information indicating a confirmation instruction from the input device 1 without passing through the control unit 3, or the conversion candidate or conversion result is input. Based on this, output information may be generated and displayed on the output device 2.

  The kana-kanji conversion-prediction device 4 performs kana-kanji conversion and prediction conversion on the conversion target character string. Specifically, in accordance with an instruction from the control unit 3, a conversion candidate list for the set conversion target character string is generated. Further, the conversion target character string is converted into a kana-kanji mixed notation desired by the user by determining a conversion result based on information indicating a determination instruction for a conversion candidate input according to a user's selection operation. In the present invention, the kana-kanji conversion-prediction device 4 changes the conversion logic applied to the conversion target character string using the attribute of the word of the conversion target character string. Here, the word attribute is identification information for identifying a classification system when a group of words is classified from a certain point of view. For example, in addition to the part of speech, the attribute is personal name / place name / food or is synonymous. A set of words and antonyms, or a semantic concept as expressed in a thesaurus may be included. Note that the conversion logic is to determine what attribute (or combination thereof) the character string to be converted is to be converted. Specifically, the conversion logic includes a syntax that can be taken by the character string to be converted and This is information indicating the attribute of the word in the syntax. For example, for a conversion target character string including words that have different meanings in one reading, there is conversion logic for converting the words into words that have different meanings.

  The kana-kanji conversion-prediction device 4 specifically converts the conversion logic corresponding to the attribute of the word based on the attribute of the word taken by the input character string to be converted and the conversion result at that time. The learning is performed by increasing the priority of the conversion logic used to extract the conversion result. In addition, the kana-kanji conversion-prediction device 4 performs each conversion logic learned for a newly input target character string in a combination of word attributes (or the closest combination of attributes) that the conversion target character string can take. Based on the priority, the priority order is determined in the conversion logic applicable to the character string to be converted, and the conversion process is performed according to the priority order.

  In the present embodiment, the kana-kanji conversion-prediction device 4 includes a kana-kanji conversion-prediction control unit 41, a kana-kanji conversion-prediction engine (core unit) 42, and a kana-kanji conversion-prediction engine (user-dependent). Part) 43 and a kana-kanji conversion-prediction dictionary 44.

  The kana-kanji conversion-prediction control unit 41 gives necessary instructions to the kana-kanji conversion-prediction engine (core unit) 42 and the kana-kanji conversion-prediction engine (user-dependent unit) 43, so that the user input history (conversion) Kana-Kanji conversion-prediction engine (user part) 43 and Kana-Kanji conversion-prediction engine (core part) 42 and Kana-Kanji conversion-prediction engine (user-dependent part) 43 are integrated. To generate a conversion candidate list.

  Specifically, the kana-kanji conversion-prediction control unit 41 can apply the conversion target character string input to the kana-kanji conversion-prediction engine (user-dependent unit) 43 when the conversion result is confirmed. In the conversion logic, a learning process for increasing the priority of the conversion logic used to extract candidates determined as conversion results is performed. In addition, when a conversion target character string is set, a conversion logic that changes according to a learning process by the kana-kanji conversion-prediction engine (core unit) 43 according to the kana-kanji conversion-prediction engine (user-dependent unit) 43 is set. Based on the priority, a process for generating a conversion candidate list for the conversion target character string is performed.

  The kana-kanji conversion-prediction engine (core unit) 42 performs kana-kanji conversion processing and predictive conversion processing using the kana-kanji conversion-prediction dictionary 44, and generates a conversion candidate list for the conversion candidate character string. The kana-kanji conversion-prediction engine (core unit) 42 is a conversion candidate list for the conversion target character string according to the priority of conversion logic that changes according to the learning process by the kana-kanji conversion-prediction engine (user-dependent unit) 43. Is generated.

  The kana-kanji conversion-prediction engine (user-dependent unit) 43 is a processing unit that is adapted to the user by the user input history in the kana-kanji conversion process and the predictive conversion process. The kana-kanji conversion-prediction engine (user-dependent unit) 43 learns (changes) the priority of the conversion logic based on the user input history. The kana-kanji conversion-prediction engine (user-dependent unit) 43 describes the priority of each conversion logic in the combination of two or more conversion logics that can be applied to the same character string as information indicating the learning result. Define a priority table for The priority table may describe the priority of each conversion logic in the combination only for the predetermined conversion logic, or the conversion logic determined from the conversion target character string or the combination thereof May be added as needed to describe the priority of each conversion logic in the combination. The priority of the conversion logic may be a counter that counts up each time the conversion logic is prioritized.

  The kana-kanji conversion-prediction dictionary 44 stores each notation of a word having the reading (notation of kana-kanji combination presented as a conversion candidate) and the attribute of the word indicated by the notation in association with the reading. A plurality of word attributes may be registered according to the number of classifications that organize the words.

  The kana-kanji conversion-prediction control unit 41, the kana-kanji conversion-prediction engine (core unit) 42, and the kana-kanji conversion-prediction engine (user-dependent unit) 43 are realized by, for example, a CPU that operates according to a program. The kana-kanji conversion-prediction dictionary 44 is realized by a storage device, for example.

  Next, the operation of the present embodiment will be described with reference to FIG. FIG. 2 is a flowchart showing an operation example of the Japanese language processing apparatus according to this embodiment. As shown in FIG. 2, first, a character string to be converted is input from the user via the input device 1. For example, the input device 1 inputs a character string to be converted in accordance with a user's key operation (step A1). The control unit 3 receives the input information from the input device 1 and sets the character string indicated by the input information as a conversion target character string to the kana-kanji conversion-prediction device 4 to convert the kana-kanji conversion and predictive conversion. Instruct. For example, the control unit 3 may recognize that the space key or the like has been pressed, and notify the kana-kanji conversion-prediction device 4 of the character string input before that time as the conversion target character string.

  When the conversion target character string is set, the kana-kanji conversion-prediction control unit 41 of the kana-kanji conversion-prediction device 4 sends a conversion candidate list for the conversion target character string to the kana-kanji conversion-prediction engine (core unit) 42. Generate.

  The kana-kanji conversion-prediction engine (core unit) 42 receives the instruction from the kana-kanji conversion-prediction control unit 41, determines the priority of conversion logic applicable to the conversion target character string, and according to the priority, A conversion candidate list for the conversion target character string is generated using the kana-kanji conversion-prediction dictionary 44 (step A2). The conversion logic applicable to the conversion target character string can be specified by determining the syntax that the conversion target character string can have and the attributes that the word can have in the syntax by performing syntax analysis or checking against the dictionary. Good. Then, the conversion logics may be ranked in descending order of priority of the conversion logics in the applicable combination of conversion logics.

  The kana-kanji conversion-prediction engine (core unit) 42 is, for example, in the priority table managed by the kana-kanji conversion-prediction engine (user-dependent unit) 43, the same combination as the combination of conversion logics applicable to the conversion target character string. Are described, the conversion logics may be ranked according to the priority of each conversion logic in the combination. Also, for example, if the same combination is not described in the priority table, the best matching combination is selected from the described combinations, and the conversion logic is ranked according to the priority of each conversion logic in the combination. Just do it. In addition, when the corresponding combination is not described in the priority table, each conversion logic is ranked according to a predetermined rule (the order in which the conversion logic is specified or the registration order in the dictionary of words having the attribute). Just add. Or, after ranking each conversion logic that can be applied based on a predetermined priority, according to the priority of each conversion logic in the most matching combination described in the priority table The order of the conversion logic may be changed.

  Here, a case where the character string “Kate” is input as a conversion target will be described as an example. At this point, nothing is described in the priority table. The kana-kanji conversion-prediction engine (core unit) 42 performs a syntax analysis on the conversion target character string “Kate” or compares it with a dictionary, so that the conversion target character string is a single word “Kate”. Conversion logic that can be applied to the conversion target character string “Kate” by determining that the word “Kate” can have the attribute “verb combined form + auxiliary verb” and “noun”. The two conversion logics are specified: conversion logic 1 indicating that “katte” is converted into “verb consecutive form + auxiliary verb“ te ”” and conversion logic 2 indicating that conversion is performed with “noun”. In the priority table managed by the Kanji conversion-prediction engine (user-dependent unit) 43, the priority of the conversion logic in the combination of the conversion logic 1 and the conversion logic 2 is described. Here, since nothing is described in the priority table, the kana-kanji conversion-prediction engine (core unit) 42 uses the conversion logic 1 and the conversion logic according to a predetermined rule. 2. Here, it is assumed that a ranking of conversion logic 1> conversion logic 2 is performed.

  The priority order can also be determined by the kana-kanji conversion-prediction engine (user-dependent unit) 43 in response to an instruction from the kana-kanji conversion-prediction engine (core unit) 42. When the priority of conversion logic applicable to the conversion target character string is determined, the kana-kanji conversion-prediction engine (core unit) 42 generates a conversion candidate list for the input conversion target character string according to the determined priority. .

  For example, the kana-kanji conversion-prediction engine (core unit) 42 applies the conversion logic to the conversion target character string in order from the conversion logic with the highest priority, and specifies the attribute of each word according to the syntax indicated by the conversion logic. Then, for each word, conversion candidates are presented in an order according to the priority of conversion logic by extracting from the dictionary the notation (Kana-Kanji mixed notation) whose reading and attribute match, and listing them as conversion candidates. A conversion candidate list may be generated. For each conversion logic, the number of cases may be limited such that a predetermined number of cases is extracted in the order of dictionary registration.

  For example, in the Kana-Kanji conversion-prediction dictionary 44, two words, “buyet (verb consecutive form“ buy ”+ auxiliary verb“ te ”) and“ katte (noun) ”, are registered as kana-kanji mixed notation that reads“ katte ”. Suppose that it was done. The kana-kanji conversion-prediction engine (core unit) 42 converts conversion logic 1 (“verb conjunctive form + auxiliary verb“ te ””)> conversion logic 2 (“noun”) with respect to conversion logic applicable to the conversion target character string “KATE”. When the priority order “”) is determined, conversion candidate 1 “buy” is first extracted by applying conversion logic 1 (“verb consecutive form + auxiliary verb“ te ””). Next, the conversion logic 2 (“noun”) is applied to extract the conversion candidate “Kute”. And 1. "Buy", 2. A conversion candidate list having a presentation order of “independent” is generated. When a plurality of notations can be extracted with one conversion logic, the presentation order may be determined according to a predetermined rule (dictionary registration order, etc.).

  Then, the generated conversion candidate list is displayed. Specifically, the control unit 3 acquires a conversion candidate list generated by the kana-kanji conversion-prediction engine (core unit) 42 via the kana-kanji conversion-prediction control unit 41 as information on conversion candidates, Output information for displaying the conversion candidate list is generated and displayed on the output device 2.

  The user selects and determines a conversion candidate to be input from the conversion candidate list displayed by the output device 2. The input device 1 inputs a confirmation instruction for one of the conversion candidates according to the user's selection operation (step A3). Upon receiving the input information from the input device 1, the control unit 3 confirms the conversion candidate instructed as a conversion result, and notifies the kana-kanji conversion-prediction device 4 of the determined conversion candidate.

  The kana-kanji conversion-prediction control unit 41 of the kana-kanji conversion-prediction device 4 receives the notification of the confirmed conversion candidate and can determine the kana-kanji conversion-prediction engine (user-dependent unit) 43 based on the conversion result. The conversion logic to be performed is reflected as a learning result (step A4).

  For example, the kana-kanji conversion-prediction control unit 41 notifies the conversion candidate determined as the conversion result to the kana-kanji conversion-prediction engine (user-dependent unit) 43, and the kana-kanji conversion-prediction engine (user-dependent unit) 43 The priority of the conversion logic is updated based on the conversion result and the attribute of the word included in the conversion target character string. Specifically, for the combination of conversion logics applicable to the conversion target character string, the priority of the conversion logic used for extracting conversion candidates determined as the conversion result is increased. For example, the kana-kanji conversion-prediction engine (user-dependent unit) 43 selects a fixed conversion candidate among the conversion logics included in the priority table defined for the combination of conversion logics applicable to the conversion target character string. A conversion priority counter indicating the priority of the conversion logic used for extraction is counted up. If the priority table does not define a conversion logic combination applicable to the conversion target character string, define the combination and initialize the priority of each conversion logic in the combination. Then, the conversion priority counter indicating the priority of the conversion logic used for extracting the determined conversion candidate may be counted up.

  In addition, for the combination of conversion logics applicable to the conversion target character string, increasing the priority of the conversion logic used for extracting the conversion candidate determined as the conversion result can be taken by the conversion target character string. It means that the priority of the attribute of the word of the notation determined as the conversion result is raised with respect to the syntax and the attribute that the word can have in the syntax.

  FIG. 3 is an explanatory diagram illustrating an example of a priority table. As shown in FIG. 3, for a combination of two or more conversion logics that can be applied to the same character string, for each conversion logic included in the combination, an identifier for identifying the conversion logic and the contents of the conversion logic And a conversion priority counter indicating the priority of the conversion logic in the combination.

  In FIG. 3, conversion logic 1 indicating conversion of a conversion target character string with “verb combined form + auxiliary verb“ te ”” and conversion logic indicating conversion with “noun”, as applied to “Kate”. 3 shows an example of a priority table defined for a combination with 2. In Fig. 3, a conversion target character string to which conversion logic 1 and conversion logic 2 can be applied is extracted using conversion logic 2. Also shown is an example of updating the conversion priority counter when a conversion candidate is confirmed, as shown in Fig. 3. For example, when the conversion priority counter of each conversion logic is initialized to 0, For the conversion target character string. "Buy", 2. When “unknown” is presented as a conversion candidate list and “unknown (noun)” is confirmed as a conversion result, the conversion priority counter of the conversion logic 2 is updated from 0 to 1.

  By these steps, when the next character string is input, the conversion logic combination corresponding to the attribute of the word that the character string can take is converted according to the priority order of the conversion logic reflecting the conversion result for the previous input character string. By extracting the conversion candidates, the user can adapt the presentation order of the conversion candidates. For example, for the character string “Kate”, 1. “Buy” and 2. 1. From the conversion candidate list that presents “selfish”; In the case where the combination of conversion logic 1 (“verb consecutive form + auxiliary verb“ te ””) and conversion logic 2 (“noun”) is combined with “conversion” as the result of conversion, conversion logic 2 is given priority. Next, when a character string “Torte” is applied, the order of presentation of the “handle (noun)” is changed to “take (verb combined form + auxiliary verb“ te ”. It can be displayed higher than “)”.

  Note that the kana-kanji conversion-prediction engine (user-dependent unit) 43, when the user has confirmed the conversion result “I want you to buy” in response to the character string input “I want you to do” in the past, In addition to learning to prioritize conversion logic 2 (“noun”) in a combination of conversion logics applicable to the character string “”, in the case of prioritizing conversion logic 1 (“verb conjunctive form + auxiliary verb“ te ””), This means that learning is not performed by separating them with “Kate”.

  This is because the priority of the conversion logic is defined in a combination applicable to the conversion target character string. In other words, the combination of conversion logic that can be applied to the character string “Kate want” and the combination of conversion logic that can be applied to the character string “Kate” are distinguished and applied to the character string “I want to do it”. This is because the priority of conversion logic (for example, “verb conjunctive form + auxiliary verb“ te ”+ auxiliary adjective“ want ””) is not updated in a possible combination of conversion logics. As a method of distinguishing combinations, a method of distinguishing all conversion target character strings as one conversion logic, or a single conversion logic including the attributes of words before and after the word based on one word. There is a way to distinguish them.

  Thus, according to the present embodiment, the conversion result for a certain input character string is learned in association with the attribute of the word that can be taken by the character string at that time. Even if it exists, based on the attribute which the word can take, the conversion logic adapted to the user can be prioritized. Therefore, if the conversion candidate list is generated in accordance with the priority of the conversion logic, even if the word has not been input by the user in the past, the display order of the conversion candidates can be adapted to the user.

  Further, according to the present embodiment, the kana-kanji conversion-prediction engine is divided into the core part and the user-dependent part, so that it can be easily mounted on a small device (such as a mobile phone) with limited hardware resources. is there. That is, it can be realized simply by cutting out the kana-kanji conversion-prediction engine (user-dependent unit) 43 for the kana-kanji conversion / prediction processing that can be realized with a small apparatus at present. Specifically, the conversion logic application order is processed in a fixed order, and a conversion priority counter is provided for each combination so that the processing priority can be changed depending on the user's usage status. It can be realized simply by taking. Also, if the hardware resource conditions are the same, the method according to the present embodiment is expected to improve the processing speed compared to a method that is not intended for implementation in a small device.

  In the above embodiment, an example of conversion logic for kana-kanji conversion is shown, but the basic operation is the same even for conversion logic for predictive conversion. In this case, specifically, the conversion logic may be information indicating the complete reading of the word including the conversion target character string and the attribute of the word. For example, when “ka” or “ka” is input, a conversion candidate list may be generated by searching the kana-kanji conversion-prediction dictionary 44 for conversion candidates of words starting from the character string. At that time, conversion candidates may be extracted from the kana-kanji conversion-prediction dictionary 44 in accordance with the priority order of the conversion logic determined by the kana-kanji conversion-prediction engine (user-dependent unit) 43.

Embodiment 2. FIG.
Next, a second embodiment of the present invention will be described with reference to the drawings. FIG. 4 is a block diagram showing a configuration example of the Japanese language processing apparatus according to the second embodiment of the present invention. As shown in FIG. 4, in the present embodiment, the kana-kanji conversion-prediction device 4 has a kana-kanji conversion-prediction dictionary (user-dependent unit) 45 as compared to the first embodiment shown in FIG. 1. It is different in point.

  The kana-kanji conversion-prediction dictionary (user-dependent unit) 45 stores information indicating the learning result of the word level in association with each reading registered in the kana-kanji conversion-prediction dictionary 44. As information indicating the learning result at the word level, for example, information indicating the determination frequency may be stored for each word having the reading (word distinguished by kana-kanji mixed notation). Moreover, you may memorize | store the information which shows the word used as the newest conversion result among the words with the reading. The kana-kanji conversion-prediction dictionary (user-dependent unit) 45 may be implemented as a single dictionary by being incorporated in the kana-kanji conversion-prediction dictionary 44.

  FIG. 5 is a flowchart showing an operation example of the Japanese language processing apparatus according to this embodiment. As shown in FIG. 5, in the present embodiment, the learning result based on the conversion candidate determination result is reflected not only at the conversion logic but also at the word level (step B4). For example, the kana-kanji conversion-prediction control unit 41 notifies the kana-kanji conversion-prediction engine (user-dependent unit) 43 of the conversion candidates determined as the conversion result. The kana-kanji conversion-prediction engine (user-dependent unit) 43 to which the conversion result is notified reflects the conversion logic to be prioritized in the priority table and the word to be prioritized for conversion candidate reading (kana-kanji mix) The kana-kanji conversion-prediction dictionary (user-dependent unit) 45 is reflected. The kana-kanji conversion-prediction engine (user-dependent part) 43 stores information indicating a word-level learning result registered in the word-kana-kanji conversion-prediction dictionary (user-dependent part) 45 in association with the reading of the conversion candidate. Update it. For example, when information indicating the confirmation frequency is registered for each word having the reading as information indicating the learning result at the word level, the confirmation frequency of the word that is the conversion candidate (Kana-Kanji mixed notation) is updated. do it. Further, for example, when information indicating a word having the latest conversion result among words having the reading is registered as information indicating a word level learning result, the conversion candidate word (kana-kanji) (Mixed notation) may be registered as the latest conversion result.

  Further, in this embodiment, when the kana-kanji conversion-prediction engine (core part) 42 generates the conversion candidate list in step B2, the priority managed by the kana-kanji conversion-prediction engine (user-dependent part) 43 is managed. Based on the priority order of each conversion logic determined by the table and the learning result of the word level stored in the kana-kanji conversion-prediction dictionary (user-dependent part) 45, the presentation order of each conversion candidate is determined.

  For example, when the kana-kanji conversion-prediction engine (core unit) 42 extracts conversion candidates according to the priority of conversion logic, the latest conversion result is obtained with respect to the presentation order of words extracted by the conversion logic. The order may be such that the words that have been confirmed or the words with a high frequency of determination are presented earlier. Depending on the learning result at the word level, the learning result at the word level may be prioritized over the priority order of the conversion logic. For example, the order of the words that are the latest conversion result or the words whose determination frequency exceeds a predetermined value may be presented early regardless of the priority of the conversion logic. Other points are the same as those in the first embodiment.

  As described above, according to the present embodiment, the conversion result for a certain input character string is learned in association with the attribute of the word taken by the character string at that time, and the reading indicated by the character string at that time is read. For example, it is possible to save a case in which rearrangement at the logic level is not appropriate for words input in the past. In other words, by providing a user-dependent part in the kana-kanji conversion-prediction dictionary, it is possible to meticulously store a case where it is desired to have an exception at the word level for adaptation in logic units.

  The present invention can be applied to any device that performs Japanese conversion processing on an input character string, and in particular, to a device that is limitedly used by a single user, such as a portable terminal or a personal computer. It can be suitably applied.

It is a block diagram which shows the structural example of the Japanese language processing apparatus by 1st Embodiment. It is a flowchart which shows the operation example of the Japanese language processing apparatus by 1st Embodiment. It is explanatory drawing which shows an example of a priority table. It is a block diagram which shows the structural example of the Japanese language processing apparatus by 2nd Embodiment. It is a flowchart which shows the operation example of the Japanese language processing apparatus by 2nd Embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Input device 2 Output device 3 Control part 4 Kana-kanji conversion-prediction apparatus 41 Kana-kanji conversion-prediction control part 42 Kana-kanji conversion-prediction engine (core part)
43 Kana-Kanji conversion-prediction engine (user dependent part)
44 Kana-Kanji conversion-predictive dictionary 45 Kana-Kanji conversion-predictive dictionary (user-dependent part)

Claims (7)

A Japanese processing device that performs Japanese conversion on the input character string to be converted,
Priority of conversion logic that defines what attribute the conversion target character string is converted to as a word based on the attribute of the word that the conversion target character string can take and the conversion result determined for the conversion target character string Japanese language processing device characterized by comprising Japanese conversion means for changing the character. Japanese conversion means
When the conversion result is determined for the conversion target character string, conversion that increases the priority of the conversion logic used to extract the notation that is the conversion result for the combination of conversion logics applicable to the conversion target character string Logic learning means,
A conversion processing unit that generates a conversion candidate list for the conversion target character string according to the priority of each conversion logic in a combination of conversion logics applicable to the conversion target character string when the conversion target character string is input; The Japanese language processing apparatus according to claim 1. The conversion logic learning means defines a logic priority table for describing the priority of each conversion logic in a combination of two or more conversion logics that can be applied to the same character string,
When the conversion result is determined for the conversion target character string, the conversion logic learning unit extracts the notation that is the conversion result in the logic priority table determined for the combination of conversion logics applicable to the conversion target character string. Updated to increase the priority of the conversion logic used for
When the conversion target character string is input, the conversion processing unit generates a conversion candidate list for the conversion target character string according to the priority of each conversion logic indicated in the logic priority table. The Japanese processing device described. The Japanese conversion means includes, as a kana-kanji conversion dictionary, kana-kanji conversion dictionary storage means for storing a reading, each notation of the word having the reading, and an attribute of the word indicated by the notation,
The conversion processing means applies the conversion logic to the conversion target character string in descending order of priority of the conversion logic, and reads from the kana-kanji conversion dictionary the notation that the reading and the attribute of the word specified by the applied conversion logic match. The Japanese language processing apparatus according to any one of claims 1 to 3, wherein a conversion candidate list is generated by extracting as conversion candidates. The conversion logic learning means defines a word priority table for describing the priority of each word indicated by the notation having the reading for the reading registered in the kana-kanji conversion dictionary,
When the conversion result is determined for the conversion target character string, the conversion logic learning means updates the word priority table defined for reading the word of the conversion result to increase the priority of the word indicated by the conversion result notation. And
When the conversion target character string is input, the conversion processing means, based on the priority of each conversion logic indicated in the logic priority table and the priority of each word indicated in the word priority table, The Japanese processing apparatus according to claim 1, wherein a conversion candidate list for a character string is generated. The Japanese conversion means is a Japanese processing method for performing Japanese conversion on the input conversion target character string,
Priority of conversion logic that defines what attribute the conversion target character string is converted to as a word based on the attribute of the word that the conversion target character string can take and the conversion result determined for the conversion target character string Japanese processing method characterized by changing A Japanese processing program for performing Japanese conversion on the input conversion target character string,
On the computer,
Priority of conversion logic that defines what attribute the conversion target character string is converted to as a word based on the attribute of the word that the conversion target character string can take and the conversion result determined for the conversion target character string Japanese language processing program for executing processes that change

Images