JPH0576066B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a character processing device that inputs sentences containing Kanji and Kana characters through Kana-Kanji conversion, and more particularly to a learning mechanism for Kana-Kanji conversion.

[Prior Art] Currently, character processing devices such as Japanese word processors generally input sentences containing kanji and kana using kana-kanji conversion.

Conventionally, kana-kanji conversion dictionaries were often stored in external memory (floppy disk, hard disk). However, as conversion methods such as batch conversion have been improved, the number of dictionary accesses has increased, and on the other hand, the demand for speeding up conversion has also increased, so a method of storing dictionaries in internal memory that can be accessed at high speed has been developed. It has become mainstream.

The internal memory is readable/writable and volatile (memory contents disappear when the power is turned off).
Two types of memory are commonly used: RAM, and ROM, which is non-writable but non-volatile (memory contents do not disappear even when the power is turned off).

When storing a dictionary in RAM, the dictionary must be loaded from external memory to RAM as a preparation before kana-kanji conversion, but this has the disadvantage of taking time to load, and in addition, RAM is more expensive than ROM. This also has the disadvantage of increasing costs. For this reason, it is now common to store dictionaries in ROM.

Since dictionaries are converted into ROM, the learning method is also changing from the past. Conventional learning methods used methods such as actually rearranging the ranks of words in the dictionary and updating the rank information (frequency information) in the dictionary, but the information in the ROM cannot be changed, so it is separate from the dictionary area.
Learning information must be stored in the RAM area. Therefore, several methods shown below have been proposed so far.

The first method is a method that uses learning data that describes only frequency information in association with words in a dictionary. The correspondence is determined by the number of the word from the beginning of the dictionary (sequential number). That is, the frequency information of the first word in the dictionary is stored in the first part of the learning data, and the frequency information of the second word from the beginning of the dictionary is stored in the second part of the learning data. The following information is stored in the same way.

In the second method, for the word being learned,
This method stores the address of the word in the dictionary.

The third method is to create a learning dictionary that is a subset of the real dictionary, which stores the entire word information (substantive data) in the dictionary, rather than simply storing the addresses of the words being learned, and then converts the words. This method preferentially converts words that are sometimes listed in learning dictionaries.

[Problems to be Solved by the Invention] The three learning methods listed above all have drawbacks.

A drawback of the first method and the second method is that the learning data depends on the contents of the dictionary ROM.

In the first method, when a new word is added or an existing word is deleted, subsequent learning data must be shifted.

In the second method, the situation becomes even more severe, and if there is any change in the dictionary, i.e., if a new word is added, if an existing word is deleted, or if the word information is corrected, each When the existence address of a word is incorrect, the address value must be corrected with respect to learning data regarding subsequent words.

Of course, with any method, it is possible to correct the learning data if you know what kind of corrections have been made to the dictionary contents, but generally speaking, when updating the dictionary contents, there are many places to update, so it is possible to correct the learning data. It is not realistic to know the changes in

In addition, the third method does not depend on the contents of the dictionary ROM, but it retains the word information as is, so
The disadvantage is that it requires a considerable amount of memory for training data. For example, if you need 10 bytes of word information per word, the number of words being learned is 1.
Assuming there are 10,000 words, the learning data will be approximately 100K.
This method requires many bytes of memory, so this method cannot be used in reality.

In other words, conventional learning methods have the disadvantage that compatibility cannot be ensured when the contents of the dictionary change, and if compatibility is to be ensured, a huge amount of memory for learning data is required.

[Means for Solving the Problems (and Effects)] In order to solve the above problems, according to the present invention, a character processing device is provided with an input means for inputting a kana character string and a word pronunciation. , a dictionary means in which word information including notation is stored in correspondence, and the dictionary means to convert the kana character string inputted from the input means into the notation of the word whose reading is the kana character string. A conversion means for converting, a selection means for selecting one word in response to a first instruction from the words converted by the conversion means, and a pointer in the dictionary means of the word selected by the selection means is stored. and a pointer in the dictionary means of the word whose reading is the inputted kana character string during conversion by the conversion means, and a pointer that matches the obtained pointer is selected as the learning word pointer storage means. a first control means that determines whether the word pointer is stored in the word pointer storage means, and when it is determined that the word pointer is stored, gives priority to the word corresponding to the pointer as a conversion result; , storage control means for extracting word information stored at the position of the pointer stored in the learning word pointer storage means from the dictionary means and storing the extracted word information in the external storage means; , searches the dictionary means for a word that matches a word stored in the external storage means, obtains a pointer on the dictionary means of the word that matches through the search, and uses the obtained pointer to store the learned word pointer storage means. By including a second control means for updating the stored contents of the words, the previously selected word can be converted with priority even if the contents of the dictionary referred to at the time of conversion are different.

[Example] The present invention will be described in detail below with reference to the drawings.

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

In the illustrated configuration, the CPU is a microprocessor that performs arithmetic operations for character processing, logical judgments, etc. Control components.

Address bus AB transfers address signals indicating the components to be controlled by the microprocessor CPU. The control bus CB transfers and applies control signals for each component to be controlled by the microprocessor CPU. The data bus DB transfers data between each component.

Next, the ROM is a read-only fixed memory, and stores the control procedure by the microprocessor CPU, which will be described later with reference to FIGS. 10 to 16, and the kana-kanji conversion dictionary DIC.

Further, the RAM is a writable random access memory having a configuration of 1 word and 16 bits, and is used for temporary storage of various data from each component.
IBUF is an input buffer, and input key data from the keyboard is temporarily stored. TBUF is a text buffer in which text data being input and edited is stored. DOBUF is a homophone buffer, and if a text buffer TBUF has a homophone candidate for a character to be stored, that candidate is stored. MLEARN is memory learning data,
Among the words in the Kana-Kanji conversion dictionary DIC, the addresses of the learned words are memorized. JCTBL
is an independent word candidate table, which is a table for temporarily storing the location of independent words necessary during kana-kanji conversion and whether or not they have been learned.

KB is a keyboard that instructs various functions for this character processing device, such as character symbol input keys such as alphanumeric key, hirakana key, and katakana key, as well as conversion key, next candidate key, selection key, and learning save key. It is equipped with various function keys for various functions.

DISK is text data, disk learning data
This is external memory for storing DLEARN.
Text data and disk learning data are saved as needed, and the saved data can be called up when needed by instructions from the keyboard.

CR is the cursor register. By CPU,
Can read and write the contents of the cursor register. A CRT controller CRTC, which will be described later, displays a cursor at a position on the display device CRT corresponding to the address stored here.

DBUF is display buffer memory that stores data patterns to be displayed. When displaying text contents, a display pattern is created in the DBUF according to the contents of the text buffer TBUF.

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

Further, a CRT is a display device using a cathode ray tube or the like, and a CRT controller controls the dot-configured display pattern and cursor display on the display device CRT.

Furthermore, CG is a character generator that stores patterns of characters and symbols to be displayed on the display device CRT.

The character processing device of the present invention, which is composed of each of these components, operates in response to various inputs from the keyboard KB, and when input from the keyboard KB is supplied, an interrupt signal is first sent to the microprocessor. The microprocessor CPU reads various control signals stored in the ROM, and performs various controls in accordance with these control signals.

FIG. 2 is a diagram showing an example of a conversion operation performed by the apparatus of the present invention. 2-1 shows the initial screen. 2-2 shows the screen when the reading sequence "Kikai" is input. The cursor is displayed next to the input reading sequence. If you press the conversion key here, the screen 2-3 will appear. In 2-3, the reading sequence ``kikai'' is converted to ``machine.'' This is not the desired conversion, so when you press the next candidate key, 2-4
The screen will appear. Here, conversion candidates for "Kikai" are displayed. The first candidate is "machine" and the second candidate is "opportunity." The current candidate is the second candidate "Opportunity", and 2 is highlighted. If you press the selection key here, the screen 2-5 will appear. The second candidate "opportunity" is determined and stored in the text data, and the "opportunity" is learned at the same time. The learning results are stored in learning data on memory. If you enter "Kikai" again, the screen 2-6 will appear. When the conversion key is further pressed, the screen 2-7 appears, and since "opportunity" has been learned this time, "opportunity" is displayed as the first candidate.

In the previous operations, the learning results are only reflected in memory, and in order to save them, it is necessary to press the learning save key. The learning results will then be stored in external memory.

FIG. 3 is a diagram showing the configuration of the input buffer IBUF.

The input buffer is a buffer that stores key data input from the keyboard. The first two bytes are the buffer length, and the length of valid key data stored in the input buffer is stored as shown in the figure. Key data is stored starting from the third byte. Key data is stored as 2 bytes per character. For example, JIS X 0208 code is used. The "/" at the end of the input in the figure means that the conversion key has been pressed.

FIG. 4 is a diagram showing the structure of the text buffer TBUF.

A text box is composed of a string of characters,
Each character consists of 2 bytes. MSB of each character
is a homophone flag, where 0 means a normal character and 1 means a homophone. The remaining 15 bits represent the character code for normal characters, and the homonym number for homonyms. For example, the character code is JIS
Use the 0208 code. The homophone number is a number indicating which homophone it is on the homophone buffer DOBUF shown in FIG.

FIG. 5 is a diagram showing the structure of the homophone buffer DOBUF. Each homophone is identified by a homophone number.

Each homophone consists of "pronunciation,""current candidate number," and "i-th candidate information."

"Yomi" stores the reading of its homophone.

The "current candidate number" stores the currently displayed candidate number of the homophone. The first value is the initial value immediately after conversion.
Since candidates are displayed, "1" is stored.

The "i-th candidate information" stores the notation of each candidate, the existence address of an independent word for learning, and the like.

FIG. 6 is a diagram showing the configuration of the dictionary DIC. The dictionary exists on ROM. Therefore, the contents cannot be changed.

Each word data consists of "pronunciation", "spelling", and "part of speech".

"Yomi" is the reading information of the word, for example, "machine"
If so, ``kikai'' will be memorized. The code is JIS
The lower byte of the X0208 code is used, and each character is stored as one byte.

"Notation" is the notation information of the word, for example, "machine"
In this case, the word "machine" is stored as 2 bytes per character using JIS X 0208 code, etc.

"Part of speech" stores the part of speech of a word, for example, if it is "machine", "noun" is stored.

Word data in the dictionary is in ascending order of pronunciation (lexicographical arrangement)
It is stored in .

FIG. 6 is a diagram showing the structure of the disk learning data DLEARN.

Disk learning data is stored in external memory. When the power is turned on, it is loaded from the external memory to the work area on the RAM and used to create the memory learning data shown in FIG. When the creation of memory learning data is completed, the loaded work area is released. In other words, the disk learning data exists in RAM for only a moment.

As shown in the figure, the actual data of the word being learned consists of three fields: "pronunciation,""orthography," and "part of speech."

Learning words should normally exist in the dictionary DIC, and when they do exist, the pronunciation, spelling, and part of speech in the dictionary DIC will be memorized as they are.

In the figure, learning word 1 is "opportunity" and learning word 2 is "proofreading."

Dictionary ROM version up (function up)
In some cases, a word that does not exist in the dictionary DIC is written as a learning word, such as when the deleted word does not exist in the dictionary DIC, but in that case, the deleted word that does not exist is not reflected in the memory learning data.

FIG. 8 is a diagram showing the detailed structure of the memory learning data MLEARN.

The memory learning data stores, for each learning word stored in the disk learning data, where that word exists in the dictionary DIC.

For example, learning word 1 is "opportunity" according to Figure 7.
Therefore, the address where "opportunity" exists in the dictionary DIC is stored as the first entry. Similarly, for learning word 2, the address where "proofread" exists is stored as the second entry.

Memory learning data is data that exists to reduce the memory used for learning data, and even if this data is not available, it can be recreated at any time from disk learning data and dictionary DIC.

In this way, the memory learning data is data that depends on the version of the dictionary DIC. It is necessary to recreate the dictionary according to DIC.

FIG. 9 is a diagram showing the structure of the independent word candidate table JCTBL.

The independent word candidate table is a table that temporarily stores the words needed to analyze the input pronunciation sequence in the process of kana-kanji conversion, where they exist in the dictionary DIC, or whether they have been learned. .

For example, if the input pronunciation sequence is ``Kikaishinkoukaikakan,'' then ``Ki'' is used for analysis.
Words such as "Kika", "Kikai", "Ka", "Kai", "Kaishi", "Kaishin" are required, and the location of these words in the dictionary DIC is the field "Pointer".
is set to

Also, based on the memory learning data MLEARN, whether or not learning has been performed is set in the bit learning flag. "1" is stored when the word has been learned, "0" when it has not been learned.
is stored.

In the case of the figure, it is assumed that only "opportunity" and "passion" have been learned, and "1" is stored in the learning flag. "0" is stored in other words.

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

FIG. 10 is a flowchart of the part in which key input is received and processed.

Step 10-1 is a memory learning data creation process, in which the memory learning data is initialized as shown in FIG. This process is normally executed only once immediately after the power is turned on.

Step 10-2 is a process of taking in data from the keyboard. In step 10-3, the type of key taken in is determined, and the process branches to a processing routine for each key.

When the conversion key is input, the process branches to step 10-4, and in step 10-4, conversion processing for kana-kanji conversion is performed as detailed in FIG. Thereafter, the process branches to display processing in step 10-9.

When the next candidate key is input, step 10-5
Then, in step 10-5, the next candidate process detailed in FIG. 14 is performed. Then step
The process branches to display processing 10-9.

When the selection key is input, the process branches to step 10-6, and in step 10-6 the selection process detailed in FIG. 15 is performed. Then step 10-
The process branches to display processing in step 9.

When the learning save key is input, step 10-
The process branches to step 7, and in step 10-7, learning and storage processing detailed in FIG. 16 is performed. Thereafter, the process branches to display processing in step 10-9.

If any other key is pressed, the process branches to step 10-8, and other processing such as insertion, deletion, etc. performed in a normal character processing device is performed. Thereafter, the process branches to display processing in step 10-9.

In step 10-9, the processing results up to now are displayed, and then the process loops to 10-2 to receive key input again.

FIG. 11 is a detailed flowchart of the "memory learning data creation process" in step 10-1.

In step 11-1, first create the work area necessary to load the disk learning data in RAM.
Secure on top.

In step 11-2, the disk learning data is loaded from the external storage into the secured work area.

In step 11-3, memory learning data is initialized.

In step 11-4, one learning word is extracted from the disk learning data on the work area.

In step 11-5, it is determined whether processing has been completed for all the learning words, and if the processing has been completed, the process branches to work area release processing in step 11-8. If processing is not completed, step
Proceed to step 11-6 to determine the learning word address.

In step 11-6, a search is made to see where in the dictionary DIC the same word as the extracted learning word exists, and its address is obtained.

In step 11-7, the address obtained above is set in the memory learning data address section. dictionary
If it does not exist on the DIC, no settings will be made.

Then, since the next learning word must be processed, the process branches to step 11-4.

In step 11-8, the work area reserved for disk learning data is released and the process returns.

FIG. 12 is a detailed flowchart of the "conversion process" in step 10-4.

In step 12-1, a word search process detailed in FIG. 13 is performed to register words necessary for analyzing the input pronunciation sequence in the independent word candidate table JCTBL.

In step 12-2, the input pronunciation sequence is analyzed by morphological analysis, syntactic analysis, etc., and clause candidates are created.

In step 12-3, the likelihood of each clause candidate is calculated to determine which clause is most likely to be converted, and is determined as the first candidate. At this time, words whose learning flags are ON in the independent word candidate table are advantageous in terms of likelihood and are likely to be converted as first candidates.

In step 12-4, a homophone buffer is created based on the determined first candidate, and the conversion result is output to a text buffer.

FIG. 13 is a detailed flowchart of the "word search process" in step 12-1.

In step 13-1, one pronunciation (pronunciation to be searched) of a word necessary for analysis of the input pronunciation sequence is determined from the dictionary DIC.

In step 13-2, it is determined whether there are no more readings to be searched, and if there are no more readings to be searched, the process returns.

In step 13-3, the dictionary DIC is actually searched for the reading to be searched, and the address is obtained.

In step 13-4, the memory learning data is searched to determine whether the address found is written in the memory learning data.

In step 13-5, it is determined whether a matching address is found, and if so, the word is considered to have been learned, and step 13-5 is executed.
In step 13-6, the learning flag of the independent word candidate table is set to "1", and if it does not exist, it means that it has not been learned, so in step 13-7, the learning flag is set to "0". Thereafter, the process branches to step 13-1 to proceed to the next search reading process.

FIG. 14 is a detailed flowchart of the "next candidate processing" in step 10-5.

In step 14-1, the homophone number of the homophone to be viewed as the next candidate is obtained from the text buffer TBUF.

In step 14-2, the position of the homophone buffer is determined from the homophone number, and the current candidate number is counted up.

FIG. 15 is a detailed flowchart of the "selection process" in step 10-6.

In step 15-1, the homophone number of the homophone to be selected is obtained from the text buffer TBUF.

In step 15-2, the position of the homophone buffer is determined from the homophone number, the notation is extracted from the candidate information indicated by the current candidate number, and the text buffer is
Set as a fixed character in TBUF.

In step 15-3, the address where the independent word exists is similarly determined from the candidate information indicated by the current candidate number, and is registered in the memory learning data MLEARN.

Figure 16 shows step 10-7 "Learning storage process"
This is a detailed flowchart.

In step 16-1, a work area necessary for disk learning data is first secured on the RAM.

In step 16-2, the work area is initialized as disk learning data.

In step 16-3, one learning word pointer is taken out from the memory learning data.

In step 16-4, the dictionary DIC is referred to based on the retrieved learning word pointer, and substantive data (pronunciation, spelling, part of speech) that is word information at that position is retrieved.

In step 16-5, the extracted entity data is registered in the disk learning data set on the work area.

In step 16-6, it is determined whether or not the processing has been completed for all the learning words. If not, the looper returns to step 16-1, and the processing up to now is repeated. If the process has ended, the process branches to step 16-7.

In step 16-7, the disk learning data on the work area is saved in an external memory.

In step 16-8, the work area reserved for disk learning data is released, and the process is terminated.

[Other Embodiments] In the above description, the case where a ROM is used as the memory in which the dictionary is stored has been explained, but the situation is the same as long as it is a non-writable memory, and the present invention is applicable. For example, optical disks, CDROMs, etc. cannot have learning data (frequency data) that directly correspond to words because they cannot be written to, but the principles of the present invention can create learning data that does not depend on dictionary versions. can do.

In addition, although we have adopted an address as the way to hold the memory learning data, if it is held in a way that depends on the dictionary contents instead of requiring less memory capacity,
The same is true no matter how you hold it. For example, processing can be performed with a similar configuration even if the frequency information is stored in correspondence with the consecutive numbers of words from the beginning of the dictionary.

Although the disk learning data is configured to have all the word information (substantive data), it can also be configured to have the minimum amount of information needed to identify the words in the dictionary. In that case as well, the situation is the same since the amount of memory used is large. For example, if you have only "pronunciation" and "notation" as word information, you can usually identify the word. Therefore, the disk learning data may be configured to store only "pronunciation" and "notation".

[Effects of the Invention] As explained above, according to the present invention, at the time of startup, the pointer on the dictionary of the learning word is obtained based on the entity data of the learning word stored externally, and the kana-kanji conversion is performed. In some cases, by prioritizing and converting the word corresponding to this pointer, it is possible to realize learning that does not depend on the contents of the dictionary with less memory usage.

[Brief explanation of the drawing]

Figure 1 is a block diagram of the overall configuration of the present invention, Figure 2 is a block diagram of the overall configuration of the present invention.
The figure shows an example of the operation of kana-kanji conversion in the present invention, and Figure 3 shows the input buffer in the present invention.
A diagram showing the configuration of the IBUF, FIG. 4 is a diagram showing the configuration of the text buffer TBUF in the present invention,
Figure 5 shows the homophone word DOBUF in the present invention.
FIG. 6 is a diagram showing the structure of the dictionary in the present invention.
A diagram showing the DIC configuration, FIG. 7 is a diagram showing the configuration of the disk learning data DLEARN in the present invention,
Figure 8 shows memory learning data in the present invention.
FIG. 9 is a diagram showing the structure of the independent word candidate table JCTBL in the present invention, and FIGS. 10 to 16 are flowcharts showing the operation of the character processing device of the present invention. DISK...external memory, CPU...microprocessor, ROM...read-only memory, RAM...
...Random access memory, DIC...Kana-kanji conversion dictionary, IBUF...Input buffer, TBUF...
Text buffer, DOBUF...Homophone buffer, MLEARN...Memory learning data, JCTBL
...Independent word candidate table, DLEARN...Disk learning data.

Claims (1)

[Scope of Claims] 1. An input means for inputting a kana character string; a dictionary means in which word information including pronunciations and spellings of words are stored in correspondence; a conversion means for converting a kana character string inputted from the input means into a word notation with the reading of the kana character string; a selection means for selecting a word; learning word pointer storage means for storing a pointer in the dictionary means of the word selected by the selection means; It is determined whether a pointer that matches the found pointer is stored in the learning word pointer storage means, and if it is determined that a pointer that matches the found pointer is stored in the learning word pointer storage means, the pointer is a first control means for prioritizing a word corresponding to a word as a conversion result; and upon receiving a second instruction, extracts, from the dictionary means, word information stored at the position of the pointer stored in the learning word pointer storage means. storage control means for storing the extracted word information in the external storage means; upon activation, searching the dictionary means for words that match the words stored in the external storage means; A character processing device comprising: a second control means for determining a pointer of a word on the dictionary means, and updating the storage contents of the learning word pointer storage means using the determined pointer.