JP2756252B2 - Character processor - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention relates to a character processing device that determines a first candidate by referring to an example when converting a kana character string representing a reading into a notation and outputting the notation. It is about the method.

[Conventional technology]

2. Description of the Related Art Conventionally, as a character processing device for converting a character string, for example, a kana-kanji conversion device in a Japanese word processor has been widely used. As a result of kana-to-kanji conversion, polysemy disambiguation has been performed for a large number of homonym candidates due to the grammatical characteristics of the part of speech of each word, but words and words co-occur semantically. Since no consideration was given to the determination of the homophone pair, erroneous conversion was often performed. Here, a combination of a word and a word, that is, a combination in a word pair dictionary storing an example has been used for determining a first candidate.

[Problems to be solved by the invention]

However, for all individual words, covering combinations of words is not efficient because of the large number of words, and is not realistic.

Even if a proper first conversion candidate is obtained for a registered word pair, for a non-registered word pair,
Naturally, it cannot be applied, so the same effect cannot be obtained.

[Means for solving the problem]

In order to solve the above-mentioned problems of the related art, a character processing device of the present invention includes an input unit that inputs a kana character string, a word dictionary that stores word reading, notation, and a semantic attribute in association with each other. A set of a word having a co-occurrence relationship with a word group having a certain semantic attribute and the corresponding semantic attribute, and combination information indicating an adjunct or an adjunct string that can be connected to the set in the co-occurrence relationship as an example The example dictionary, the attached word table in which the combined information is associated with the attached word or the attached word string so that a plurality of attached words or attached word strings can be associated with one combined information, and the word dictionary is referred to. A phrase candidate creating means for creating a candidate for each phrase in the phrase string reading the kana character string input from the input means, and referring to the example dictionary and the attached word table for each candidate for the phrase. , Independent phrases of each clause And the semantic attributes of the independent word candidates of the other phrases correspond to the set of examples stored in the example dictionary, and correspond to the attached word or the attached word string that joins the independent word candidates in the attached word table. Extracting means for extracting a set of clause candidates whose attached information matches the combined information of the example, and, if there is a set of clause candidates extracted by the extracting means, giving priority to the clause candidates forming the set. ,
Output means for determining a first candidate of each of the phrases and outputting a notation of the first candidate.

According to another aspect of the present invention, a character processing method includes:
An input step of inputting a kana character string, a word reading, a notation, and a phrase string in which a kana character string input from the input means is read by referring to a word dictionary stored in association with a semantic attribute. Candidate candidate creating step of creating a candidate for each phrase, a set of a word having a co-occurrence relationship with a word group having a certain semantic attribute, and an adjunct that can combine the set in the co-occurrence relationship Or, an example dictionary storing, as an example, combined information representing an attached word sequence, and associating combined information with an attached word or an attached word sequence so that a plurality of attached words or attached word sequences can be associated with one combined information item. With reference to the attached word table, one of the independent word candidates of a certain phrase and the semantic attribute of another independent word candidate correspond to a set of examples stored in the example dictionary, and Attachment to join independent word candidates Alternatively, if there is an extraction step of extracting a set of phrase candidates in which the combination information associated with the attachment word string and the attachment word table matches the combination information of the example, and if there is a set of clause candidates extracted by the extraction step, An output step of determining a first candidate of each of the phrases by giving priority to the phrase candidates forming the set, and outputting a notation of the first candidate.

As a result, for example, an example of “release an animal” is registered, and conversion can be performed by applying the above example to all words to which a semantic attribute of “animal” is given.

〔Example〕

 Hereinafter, the present invention will be described in detail with reference to the drawings.

 FIG. 1 shows an example of the overall configuration of the present invention.

In the configuration shown in the figure, a CPU is a microprocessor, performs calculations for character processing, performs logical judgment, and the like, and is connected to those components via an address bus AB, a control bus CB, and a data bus DB. Control elements.

The address bus AB transfers an address signal indicating a component to be controlled by the microprocessor CPU.
The control bus CB transfers and applies a control signal of each component to be controlled by the microprocessor CPU. The data bus DB transfers data between the components.

Next, the ROM is a fixed read-only memory. PA
This is a program area in which a control procedure and the like shown in FIGS. 6 to 14 by the microprocessor CPU described later are stored.

The RAM is a writable random access memory having a structure of one word and 16 bits, and is used for temporarily storing various data from each component.

TBUF is a document buffer, and is a memory for storing document information input from the keyboard KB.

YBUF is an input reading buffer memory for storing readings input from the keyboard KB.

 DIC is a word dictionary for performing kana-kanji conversion.

DBPOOL is a homophone candidate storage memory that stores information obtained by analyzing and converting YBUF readings into phrases.

YJPOOL is a memory that stores information on examples applied to homophone candidates in cooperation with DBPOOL.

LRNDAT is a learning data storage memory that stores learning states of individual words and examples.

FZTBL is an attached word string conversion table for causing attached words to react to binding information stored in DIC.

KB is a keyboard, and character symbol input keys such as alphabet keys, hiragana keys, katakana keys, and
Various function keys such as a conversion key for instructing conversion are provided.

In the figure, YOMI is a key for inputting readings, CON
Is a conversion instruction key for converting the input reading, NXT is a next candidate conversion instruction key for changing the conversion candidate and converting it to the next candidate, and SEL is determined as the current homophone display candidate and simultaneously displays the candidate notation. This is a selection key for instructing learning.

DISK stores documents created in a memory for storing fixed-form documents, and the stored documents are called up when necessary by keyboard instructions.

CR is a cursor register. The CPU can read and write the contents of the cursor register. A CRT controller CRTC described later displays a cursor at a position on the display device CRT corresponding to the address stored here.

DBUF is a display buffer memory for storing patterns such as document information stored in TBUF.

The CRTC plays a role of displaying the contents stored in the cursor register CR and the buffer DBUF on the display CRT.

The CRT is a display device using a cathode ray tube or the like, and the display of the dot configuration pattern and the cursor on the display device CRT is controlled by a CRT controller. Furthermore, C
G is a character generator that stores character and symbol patterns to be displayed on the display device CRT.

In the character processing apparatus of the present invention comprising such components, the apparatus operates in response to various inputs from the keyboard KB. When an input from the keyboard KB is supplied, first, an interrupt signal is generated by the microprocessor CPU.
The microprocessor CPU reads various control signals stored in the ROM, and performs various controls according to the control signals.

FIG. 2 is a diagram showing a configuration of a word dictionary DIC used for kana-kanji conversion.

The dictionary is composed of word information used for notation conversion and example information for determining a first candidate at the time of notation conversion.
The word information is composed of a word ID unique to the heading word, a reading that stores the reading of the word, a notation that stores the notation of the word, a part of speech that stores the part of speech of the word, and a meaning that stores the semantic attribute of the word. The word ID is a serial number with the first word being 1. The example information is the number of example information that the headword has, the example ID unique to the example, the type that describes the example type whether the example is an individual example or a semantic example, and the example type if the example type is an individual example. If the word ID of the partner's word is stored and if it is a semantic example, an example that stores the semantic attribute that is the partner of the example. It consists of the stored coupling information and the reversal storing information on whether or not the example can be used even if the example is reversed. For a word having no example, 0 is stored in the number of example information. example
The ID is a serial number with the first example being 1 as in the case of the word ID.

In the figure, for example, the word "buy" is distinguished by a serial number of 300, and has word information that the reading is "ka", the notation is "buy", and the part of speech is "verb". It has two separate word examples, "buy sake" and "buy quarrel." Also, the word "salmon" is distinguished by a consecutive number of 500, the word is "salmon", the notation is "salmon", the part of speech is "noun", and the semantic attribute is "animal". Since 0 is stored in the number of examples, no example in which “salmon” is a headword is registered.

Fig. 3 shows a homophone pool DBPOOL and an example information pool
It is a figure showing the composition of YJPOOL. (A) is an entire configuration of a homophone pool. The homophone pool is composed of each homophone information created by the kana-kanji conversion process. (B)
FIG. 3 is a diagram showing a configuration of homophone information. Homophone information is
A homophone ID which is a serial number for distinguishing the homophone from other homophones, a data length indicating a used memory size of the homophone information, and a candidate number for storing the number of stored homophone candidates. , Candidate information storing information on homophone candidates, and a priority number storing the number of priority candidates described later among candidate information stored in the homophone. In such a homophone configuration, the candidate information at the head of the column of candidate information is assumed to be the first candidate to be output as a conversion result, and the number of pieces of candidate information indicated by the priority number from the head of the column will be described later. It shall be a priority candidate. (C) is a diagram showing a configuration of candidate information. The candidate information is composed of a phrase of the candidate, its word ID, a phrase end utilization for storing a phrase end utilization form, and an example information pointer for storing a pointer to example information to be described later. When the example is not applied to the candidate information and the example information does not exist, the invalid value -1 is stored in the example information pointer. (D) is the entire configuration of the example information pool YJPOOL.
The example information pool is composed of example information for storing information on examples applied to each homophone candidate. (E) is a diagram showing the configuration of example information. The example information stores the pair homophone ID that stores the ID of the homophone that is the partner of the example, the example ID that stores the ID of the applied example, and the type of the applied example (individual example or semantic example). And an example information pointer in which a pointer to another example information applied to the same candidate information is stored. Thus, the example information has a list structure. If the next example information does not exist, an invalid value -1 is stored in the example information pointer.

In the configuration of the homophone pool and the example information pool as described above, when an example is applied to a certain candidate of a certain homonym, the example information is stored in the candidate information, and the homonym that is a partner of the example is stored. It is assumed that the candidate information for a candidate also stores example information in which information of the same example is stored.

FIG. 4 is a diagram showing a configuration of an attached word string conversion table FZTBL. The attached word string conversion table is a correspondence conversion table for associating an attached word string constituting a phrase with combination information in the example information of the word dictionary DIC. The attached word reading indicates the reading of the attached word string, and the corresponding combined information indicates combined information in the example information of the dictionary in which the attached word string can be replaced.

For example, the ancillary word "kana" is combined information "ga" or "wo", the ancillary word string "Saemoga" is combined information "ga",
Each of them can be replaced to determine the binding.

FIG. 5 is a diagram showing a configuration of the learning data LRNDAT. (A) is an entire configuration of learning data. The learning data is composed of two pieces of word learning data and example learning data. The pointer to the word learning stores an index to the head of the word learning string in the learning data, and the pointer to the example learning stores the index to the head of the example learning data string in the learning data. (B) is a diagram showing a configuration of a word learning data sequence. The word learning data string is composed of the used size and each word learning data. The word learning data use size stores the use size of the word learning data sequence. The word learning data corresponds to the ID of each word, and 1 is stored for a learned word and 0 is stored for an unlearned word. (C) is a diagram showing a configuration of an example learning data sequence. The example learning data sequence is composed of the used size and each example learning data. The example learning data use size stores the use size of the example learning data sequence. The example learning data corresponds to each example ID, and 1 is stored for a suppressed use prohibited example and 0 is stored for an unrestricted use or brain example.

 The operation of the above embodiment will be described according to a flow.

FIG. 6 is a flowchart showing the operation of the character processing apparatus of the present invention.

In step S6-1, the process waits until a key is pressed from the keyboard and an interrupt occurs. When a key is input, the key is determined in S6-2, and S6-3, S6-
4. The process branches to any one of steps S6-5, S6-6, and S6-7.

S6-3 is a process when the reading input key YOMI is pressed, and the code of the pressed reading is stored in the input reading buffer memory YBUF.

S6-4 is a process performed when the conversion key CON is pressed, and converts the character string to be converted into kana to kanji, which is input in S6-3 and stored in YBUF, and outputs it to the output buffer. I do. When converting to kanji, the first candidate of the same phonetic word is determined using the example information existing in the word dictionary DIC.

S6-5 is processing when the next candidate conversion key NXT is pressed, and displays another homophone candidate in the output buffer output in S6-4.

S6-6 is processing when the selection key SEL is pressed, and determines the homophone in the output buffer displayed on the screen, and outputs the determined character string in the document. further,
A process for learning the selected word is performed.

S6-7 is processing when a key other than YOMI, CON, NXT, and SEL (for example, a key used for editing a document such as a cursor movement key) is pressed, and is generally performed in the same type of character processing apparatus. This process is well-known and will not be described in particular.

S6-8 is a display process for displaying a portion changed as a result of the above process. This is a widely-used process in which one character of data in a document is read, developed into a pattern, and output to a display buffer.

FIG. 7 is a detailed flowchart of the process in S6-4.

S7-1 is a process of analyzing a character string to be subjected to kana-kanji conversion, which is input after being divided and written in units of phrases, and outputting candidates for kana-kanji conversion output to the homophone pool. A process in which character strings are sequentially taken out in unit of division, a word dictionary is searched and analyzed, and only candidates recognized as phrases are output to a homophone pool. This is a generally performed process, and is not particularly described because it is known.

S7-2 checks whether or not the example pattern stored in the word dictionary exists for the analysis result output to the homophone pool in S7-1, and if the example pattern exists, the Pick up homophone candidates to be used as examples as priority candidates.

In step S7-3, the first candidate for kana-kanji conversion is determined from the priority candidates picked up in step S7-2 and the candidates whose words have been learned.

S7-4 is a process for displaying the output of the Kana-Kanji conversion stored in the output buffer. This process is generally performed in the same type of character processing device, and is not particularly described since it is known.

FIG. 8 is a detailed flowchart of the process of S7-2.

Homophone words to be subjected to priority candidate extraction are sequentially extracted from the beginning, and the following processing is performed on all the homophone words until there is no more target homophone word.

 In S8-1, candidate information in the homophone is extracted.

In S8-2, it is determined whether or not all candidate information has been processed, and if all have been processed, priority candidate extraction ends. If the next candidate information can be extracted, it is determined in S8-3 whether the candidate information includes example information. Whether or not the example information exists can be determined by the example information pointer in the candidate information. That is, if the value of the example information pointer is “−1”, there is no example information. Otherwise, the example information exists in the area indicated by the example information pointer. If there is no example information, the priority candidate extraction process ends.

If the example information exists, the backward example information is extracted in S8-4, and the forward example information is extracted in S8-5.

When the process in S8-5 is completed, the process loops to S8-1 to process the next candidate information.

FIG. 9 is a detailed flowchart of the process of S8-4.

In S9-1, the information of the examples existing in the word is fetched one by one from the dictionary according to the word ID stored in the candidate information.

In S9-2, it is determined whether or not an example exists in the dictionary for the word of the candidate information in S9-1, and whether all the examples in the dictionary have been taken. If an example exists
Proceeding to S9-3, if there is no example or if all the examples in the dictionary have been taken, the backward example extraction processing is finished and the routine returns.

In S9-3, it is checked whether or not the retrieved example can be applied in reverse. The information is stored in the dictionary. S9-4 if reversible
If not, go to S9-1.

In S9-4, when the retrieved example is applied in reverse, it is determined whether or not the input reading can be established. If it is possible, the process proceeds to S9-5, and if not, the process proceeds to S9-1.

In S9-5, the information of the feasible example is stored in the example information. That is, the example information is stored in the example information pool as example information, and if there is no example information in the target homophone candidate information, the example information pointer is replaced with a pointer to the newly created example information. Store. If the example information is present in the candidate information, the newly created example information is added to the end of the example information list. In addition, the example information is stored not only in the target homophone candidate information but also in the homophone candidate information to be the partner of the example.

FIG. 10 is a detailed flowchart of the process in S9-4.

In S10-1, it is checked whether or not a phrase following the target phrase exists. If a subsequent clause exists, the process proceeds to S10-2, and if not, the process proceeds to S10-7.

In S10-2, it is checked whether the ending form of the target phrase is a continuous form. If it is a continuous type, go to S10-3; if not, go to S10-7.

In S10-3, candidate information is fetched one by one from the homophone for the following phrase.

In S10-4, it is determined whether or not the candidate information has been obtained in S10-3. S10−
The process proceeds to step S5, where all candidate information has been obtained. If no candidate information has been obtained, the process proceeds to step S10-7.

In S10-5, it is determined whether the word of the candidate information obtained in S10-3 can be a partner of the example. That is, if the type of the example is an individual example, if the word ID stored in the example item of the example information in the dictionary matches the word ID of the candidate information obtained in S10-3, the partner of the example is If they do not match, it will not be the partner of the example. If the example type is a semantic example, the semantic attribute stored in the example item of the example information in the dictionary and S10-3
If the semantic attributes obtained from the word IDs of the candidate information obtained in step 1 match, it is determined that the candidate information becomes a partner of the example. If it is determined that it will be the partner of the example, the process proceeds to S10-6, and if it is determined that it will not be the partner of the example, the process proceeds to S10-3.

In S10-6, the return that the example is
Set the code.

In S10-7, a return code indicating that the example is not established is set.

FIG. 11 is a flowchart detailing the processing of S8-5.

In S11-1, the word ID stored in the candidate information
Retrieves the information of the examples existing in the word one by one from the dictionary.

In S11-2, it is determined whether or not there is an example in the dictionary for the word of the candidate information in S11-1 and whether or not all the examples in the dictionary have been taken. If an example exists, the process proceeds to S11-3. If the example does not exist, or if all the examples in the dictionary have been taken, the forward-direction example extracting process ends, and the process returns.

In S11-3, when the extracted example is applied in the forward direction, it is determined whether or not the input reading can be established. If it is possible, the process proceeds to S11-4, and if not, the process proceeds to S11-1.

In S11-4, the information of the feasible example is stored in the example information. That is, the example information is stored in the example information pool as example information, and if there is no example information in the target homophone candidate information, the example information pointer is replaced with a pointer to the newly created example information. Store. If the example information is present in the candidate information, the newly created example information is added to the end of the example information list. In addition, the example information is stored not only in the target homophone candidate information but also in the homophone candidate information to be the partner of the example.

FIG. 12 is a detailed flowchart of the process in S11-3.

In S12-1, it is checked whether a phrase before the target phrase exists. If the previous clause exists, proceed to S12-2,
If not, the process proceeds to S12-8.

In S12-2, candidate information is fetched one by one from the homophone for the previous phrase.

In S12-3, it is determined whether the candidate information has been fetched in S12-2. If the candidate information is taken, S12-
Then, the process proceeds to S12-8, where all candidate information has been obtained. If no candidate information has been obtained, the process proceeds to S12-8.

In S12-4, it is determined whether the word of the candidate information obtained in S12-2 can be a partner of the example. That is, if the type of the example is an individual example, if the word ID stored in the example item of the example information in the dictionary matches the word ID of the candidate information obtained in S12-2, the partner of the example is If they do not match, it will not be the partner of the example. If the type of the example is a semantic example, the semantic attribute stored in the example item of the example information in the dictionary and S12-2
If the semantic attributes obtained from the word IDs of the candidate information obtained in step 1 match, it is determined that the candidate information becomes a partner of the example. If it is determined that it will be the partner of the example, the process proceeds to S12-5, and if it is determined that it does not become the partner of the example, the process proceeds to S12-2.

In S12-5, the remaining reading (attached word string) of the candidate information obtained in S12-2 other than the independent part is converted into combined information by the attached word string conversion table.

In S12-6, it is determined whether or not the homophone candidate of the previous phrase satisfies the combining condition of the example. That is, in S12-5, if conversion to connection information cannot be performed by the attached word string conversion table, it is determined that the connection condition is not satisfied. If there is matching information between the combination information obtained in S12-5 and the combination information of the example information in the dictionary, it is assumed that the combining condition is satisfied. Shall not be. If the connection condition is satisfied, the process proceeds to S12-7, and if the connection condition is not satisfied, the process proceeds to S12-2.

In S10-7, the return that the example is
Set the code.

In S12-8, a return code indicating that the example is not established is set.

FIG. 13 is a detailed flowchart of the process of S7-3.

Homophones for which the first candidate must be determined are sequentially extracted from the beginning, and the following processing is performed on all the homophones until there is no more target homophone.

In S13-1, a priority candidate is picked up from candidate information in the homophone. The priority candidate pickup sets the number of pieces of candidate information having the example information pointer that is not “−1” among the pieces of candidate information included in the same word as the priority number, and sequentially rearranges the candidate information counted as the priority number to the top. When the pickup of the priority candidate is completed for all the candidate information, the process proceeds to S13-2, and a process of excluding suppressed examples from the priority candidates is performed. The priority exclusion of the suppression example is performed by performing the following processing on the candidate information of the priority number from the head of the candidate information. First, the example information pointer in the candidate information is traced, and the example ID in the example information is extracted. It is checked whether or not the example learning data indicated by the extracted example ID is in the suppression state. As a result of checking whether or not the example ID until the pointer reaches the end, that is, until the pointer value indicates “−1”, all the example information indicated by the example information pointer of the candidate information is suppressed. Only when this occurs, the candidate information is excluded from the priority candidates. That is, the priority number is reduced by 1, and the candidate information is rearranged immediately after the end of the priority candidate.

In S13-3, as a result of S13-2, it is determined whether or not the priority candidate remains based on whether or not the priority number is 0. If there is no priority candidate, the process proceeds to S13-5, where one or more priority candidates remain. If so, the flow branches to S13-4.

Of the candidate information left as priority candidates in S13-4,
Check if there is a candidate with individual word example. The existence check of the individual word example is performed by performing the following processing on the candidate information of the priority number from the head of the candidate information. First, the example information pointer in the candidate information is followed, and the type of the example is checked for all the example information indicated by the example information pointer. If there is at least one individual word example, it is determined that "an individual word example exists". Until the example information pointer reaches the end, that is, until the pointer value indicates “−1”, the type of the example is checked. If all the examples are semantic attribute examples, it is determined that there is no “individual word example”.

As a result of the determination, if there is an individual word example, the flow branches to S13-6, and if there is no individual word example, the flow branches to S13-7.

In S13-6, as a result of the individual word example search in S13-5, candidate information that is all semantic attribute examples is excluded from priority candidates.
That is, the priority number is reduced by 1, and the candidate information is rearranged immediately after the end of the priority candidate.

As a result of the above processing, it is determined in S13-7 whether the number of priority candidates has been reduced to one. That is, it is determined whether or not the priority number is one. If the priority number is 1, S13-9
Then, the priority candidate is determined as the first candidate. If the priority number is not 1, the number has not yet been reduced to one.
In 3-8, candidates are narrowed down by word learning.

In step S13-8, the word ID of the priority candidate is extracted, the word learning data indicated by the word ID is examined, and if there is a candidate whose word has been learned, the candidate whose word has been learned is determined as the first candidate in step S13-11. I do. If there are a plurality of candidates for which the word has been learned, the previously stored candidate information is set as the first candidate. If none of the candidates in the priority candidates has been word-learned, the candidate stored at the head of the priority candidates is determined as the first candidate in S13-10.

If it is determined in S13-3 that no priority candidate exists, a determination is made in S13-5 to determine a first candidate by word learning. The word ID of the candidate information is extracted, the word learning data indicated by the word ID is examined, and if there is a candidate for which the word has been learned, the candidate for which the word has been learned is determined as the first candidate in S13-11.
Determined as a candidate. If there are a plurality of candidates for which the word has been learned, the previously stored candidate information is set as the first candidate. If none of the candidate information has been word-learned, the leading candidate information is determined as the first candidate in S13-12.

FIG. 14 is a detailed flowchart of the process of S6-6.

In S14-1, it is determined whether or not an example exists in the homophone in which the candidate designated and selected by the selection key SEL is stored by referring to the example information pointer for all candidate information in the homophone. judge. That is, if the value of the example information pointer is “−1”, it is determined that there is no example. If there is no example for all candidate information, the process branches to S14-3, and if there is even one example, the process branches to S14-2.

It is determined as follows whether or not an example formed by the word selected in S14-2 exists in the example information.

First, the example information pointer in the candidate information of the selected word is followed, and the pair homophone ID and the example ID are extracted. The example information pointer of the head candidate information in the homophone indicated by the extracted pair homophone ID is followed to search whether there is any pair homophone ID that is the homophone ID in which the selected word exists. If such example information is not retrieved, it is determined that the example does not hold. Pair homophone ID is
When the example information that matches the homophone ID of the selected word is found, the example ID is taken out and it is checked whether it matches the example ID in the example information of the selected word. If the example IDs match, it is determined that an established example exists.
If they do not match, it is determined that the example does not hold. If it is determined that the example is established, the process proceeds to S14-6, and if it is determined that the example is not established, the same determination is performed for the example information next to the paired homonym. If all the examples are not satisfied for the example information of the paired homonym, the same determination processing is performed for the example information next to the selected word.
If there is no example information of the selected word, it is determined that there is no example to be established, and the flow branches to S14-5.

S14-4 is an example suppression canceling process. The example learning data indicated by the example ID of the example determined to be an established example in S14-2 is set in a usable state.

In step S14-3, when there is no example in the word selected in step S14-1, an example existence determination is performed to suppress the examples of other non-selected homonyms.

For that purpose, it is determined whether or not the example information is present in all the example information other than the selected candidate information of the homophone in which the selected word exists, and if the example exists,
In step S14-5, the example learning data indicated by the example ID is set to the use prohibited state.

In step S14-6, a word learning process is performed. The word learning data indicated by the word ID of the selected notation is set to the learning state, and the processing of canceling the learning state of the word learning data indicated by the word ID of the unselected candidate information in the homophone is performed.

The determination processing for the document of the notation selected in S14-7 is performed. The determination process is a process that is normally performed in the same type of character processing device that stores the determined character code in the document buffer TBUF, and is well known and will not be described in detail.

[Other embodiments]

Although the dictionary used in the above description has a single dictionary structure in which word information and example information are described under the same headword, the same processing is performed for each dictionary table in which word information and example information are separated. be able to. In addition, a dictionary is configured for each type of example, and each dictionary has a unique ID for the dictionary.
The process may be performed by giving the priority of the example according to the dictionary ID in which the example exists.

Further, in the above description, the example is applied only between adjacent phrases, but it is also easy to apply the example between non-adjacent phrases. At this time, by applying a syntactic analysis process (dependency analysis process) between the phrases, it is possible to avoid erroneous application of the example.

In the above explanation, the case where the example is applied between the segments input in the break-in mode has been described. By adding the processing for automatically dividing the reading sequence input in the solid-statement into the phrases, The example can be applied in the same way. It is also possible to divide the phrase into phrases while taking into account the existence of the example.

〔The invention's effect〕

As described above, according to the present invention, a set of a word group having a certain semantic attribute and a word having a co-occurrence relation and the corresponding semantic attribute, and an adjunct or an adjunct that can combine the set in the co-occurrence relation An example dictionary that stores, as an example, combination information representing a word string, and an attachment that associates a combination of information with an attachment word or an attachment word string so that a plurality of attachment words or an attachment word string can be associated with one combination information. Referring to the word table, one of the independent word candidates of a certain phrase and the semantic attribute of the independent word candidate of another phrase correspond to a set of examples stored in the example dictionary, from the candidate of each of the clauses. An attached word or an attached word string for joining independent word candidates is combined with the associated information in the attached word table to extract a set of clause candidates that match the combined information of the example. If there is a pair, The phrase candidates constituting a set by priority,
Since the first candidate of each phrase is determined, one example is stored for a group of words having the same semantic attribute that co-occurs with a certain word, so that an individual example is stored for each word in the group of words. When the notation of the first candidate word of each phrase is output without storing, a set of co-occurring words can be prioritized, so that the storage capacity required for the example dictionary can be reduced. In addition, since the same example can be used for all words having the same semantic attribute, a case occurs in which individual examples are used between words having the same semantic attribute.
There is an effect that variation in candidates due to the presence or absence of an example is eliminated. Also, when extracting a candidate corresponding to the example,
In addition to the co-occurrence relationship between one independent word candidate and another independent word candidate, the adjuncts or adjunct strings that join the independent word candidates are also considered. There is an effect that a good conversion can be realized.

Further, as an example of the example dictionary, a set of one independent word having a co-occurrence relationship and a semantic attribute of another independent word, and an adjunct or an adjunct that can join the independent words in the co-occurrence relationship By storing the combination information representing the columns, and in the auxiliary word table, it is possible to associate a plurality of additional words or additional word sequences with one combination information, and to associate the additional information with the additional words or additional word sequences, A set of independent words having a co-occurrence relation and the semantic attribute of the independent word are the same, and a plurality of sets having different adjuncts or adjunct word sequences that can combine the independent words in the co-occurrence relation are taken as one example. Since it can be stored, the storage capacity required for the example dictionary can be reduced, and if an example is registered based on a certain auxiliary word or auxiliary word sequence, the same co-occurrence can be achieved for a set of independent words and semantic attributes of the independent words. Others joining in a relationship Even if you are using the included word or attached word strings, there is an effect that can be extracted as a candidate corresponding to examples.

[Brief description of the drawings]

FIG. 1 is a block diagram of the overall configuration of the present invention, FIG. 2 is a diagram showing an example of a dictionary configuration of the present invention, FIG. 3 is a diagram showing the configurations of a homophone pool and an example information pool of the present invention, FIG. 5 is a diagram showing a configuration of an attached word conversion table of the present invention, FIG. 5 is a diagram showing a configuration of learning data of the present invention, and FIGS. 6 to 14 are flowcharts showing operations of the character processing device of the present invention. .

Claims (2)

(57) [Claims] An input means for inputting a kana character string, a word dictionary storing word reading, notation, and a semantic attribute in association with each other, a word having a co-occurrence relationship with a word group having a certain semantic attribute An example dictionary that stores, as an example, a set of a set of the set of the attribute set and the semantic attribute, and combination information indicating an adjunct or an adjunct string that can combine the set in the co-occurrence relationship. The kana character string input from the input means is read by referring to the attached word table in which the combined information is associated with the attached word or the attached word string so that the attached word string can be associated with the word dictionary. A phrase candidate creating means for creating a candidate for each phrase in the phrase sequence; and referring to the example dictionary and the adjunct word table for each of the phrase candidates, one of the independent word candidates of a certain phrase and the independence of another phrase. The meaning attribute of the word candidate is A phrase that corresponds to a set of examples stored in the example dictionary and has associated information associated with the attached word or the attached word string that joins the independent word candidates in the attached word table matches the combined information of the example. Extracting means for extracting a set of candidates; and if there is a set of clause candidates extracted by the extracting means, the first candidate of each of the clauses is determined by giving priority to the clause candidates forming the set. Output means for outputting a notation of a candidate. 2. A method of inputting a kana character string, referring to a word dictionary storing word reading, notation, and semantic attributes in association with each other, and converting the kana character string input from the input means. A phrase candidate creation step of creating a candidate for each phrase in a phrase string to be read; a group of words having a co-occurrence relationship with a group of words having a certain semantic attribute; An example dictionary that stores, as an example, linkable appendixes or link information representing linkable adjuncts, and link information and appendixes or appendixes so that a plurality of appendixes or appendix strings can be associated with one linking information. An example in which one of the independent word candidates of a certain phrase and the semantic attribute of the independent word candidate of another phrase are stored in the example dictionary with reference to the attached word table in which the columns are associated with each other. And the independent word candidate An extraction step of extracting a set of phrase candidates whose combination information associated with the attached word or the attached word string to be combined with the combination information in the attached word table matches the combination information of the example, and a phrase extracted by the extraction step If there is a set of candidates, an output step of giving priority to the phrase candidates forming the set, determining first candidates of the respective phrases, and outputting a notation of the first candidates. Method.