JPH0421160A - Document preparing device provided with learning function and its learning method - Google Patents

Abstract

PURPOSE:To prevent the generation of misconversion at the time of executing the 'KANA' (Japanese syllabary)/'KANJI' (Chinese character) conversion of a composite word by providing the document preparing device with a composite word deciding means, deciding a composite word or not a composite word and updating the frequency information of the corresponding word or composite word. CONSTITUTION:When a certain reading is inputted through a keyboard 11 and a key 11a is depressed, the reading is processed by a 'KANA'/'KANJI' conversion processing part 16 in a 'KANA'/'KANJI' control part 15. The processing part 16 decides whether the inputted reading information is a composite word or not based upon a prescribed rule. In the case of a composite word, the processing part 16 refers a composite long learning dictionary 20 and checks whether the composite word has been learned for a long time or not. When long learning has been executed, the output order of homonym candidates is determined based upon the dictionary 20. When long learning has not been executed or the word is not a composite word, the processing part 16 refers a reference long learning dictionary 19, and when the word has been learned for a long period, the output order of homonyms is determined based upon the frequency information of the dictionary 19. When the word has not been learned for a long period, the output order is determined in accordance with the frequency information of the dictionary 18. Thus, proper output order can be determined and the generation of misconversion in the 'KANA'/'KANJI' conversion can be prevented.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to a document creation device having a word and compound word learning function, particularly a long-term learning function, and a learning method thereof.

(Prior Art) Conventionally, in a document creation device, when performing homophone conversion, it is possible to refer to a plurality of words with the same pronunciation in order of frequency of use.

This is achieved by using a dictionary that stores a plurality of words with the same reading and frequency information for each word, and sequentially reading words corresponding to the input reading information from this dictionary according to the frequency information.

By the way, in a document creation device that uses such a dictionary, if the frequency information of words is fixed, even if the word is particularly used by the user, if the frequency information of that word is low, it may not appear immediately. This may cause problems such as not being able to move. Therefore, a learning function is generally used that updates the frequency information of the selected word each time the user selects one of a plurality of words with the same pronunciation. This allows the dictionary to have a word appearance frequency structure tailored to the user.

This learning function includes long-term learning and short-term learning.

In long-term learning, unless there is a specific instruction, frequency information is updated according to the frequency of word appearance. On the other hand, in short-term learning,
For example, the frequency information is initialized every time the power is turned on or every document. Here, a document creation device equipped with a long-term learning function will be described.

Conventionally, in document creation devices equipped with a long-term learning function, long-term learning has been performed on a word-by-word basis. That is, for example, if "sudden start" is inputted and "sudden start" is selected and confirmed, long-term learning is performed for "sudden" and "sudden start" respectively. For this reason, if you use this word a lot and then manually write, for example, ``kiyushakai,'' it will be converted to ``kyūshakai,'' since the frequency of ``kyū'' has increased.

On the other hand, if you use "old society" a lot, "old" becomes more frequent, and "Kiyuuhatsushin" becomes "old departure".
becomes.

(Problem to be solved by the invention) As mentioned above, conventionally, long-term learning was carried out at the word level, so when each word that has a meaning in a compound word is converted, the frequency of learning the word alone becomes effective, There was a problem that erroneous conversions were likely to occur. Note that a compound word is a word that combines two or more words to form a word with a new meaning.

The present invention has been made in view of the above-mentioned points, and an object of the present invention is to provide a document creation device and a learning method thereof that have a learning function that can prevent erroneous conversion when converting compound words into kana-kanji.

[Structure of the Invention] (Means for Solving the Problems) That is, the present invention provides a first storage means that stores a plurality of words for the same reading and frequency information of each word;
a second storage means storing a plurality of compound words corresponding to the same reading and frequency information of each compound word, and when the input reading information is not a compound word, the homophone candidate of the word corresponding to the input reading information is stored as the above-mentioned homophone candidate. If the compound word is a compound word, a homophone candidate of the compound word corresponding to the input reading information is extracted from the second storage means, and a desired homophone candidate is extracted from the extracted homophone candidates. When a homophone is selected and confirmed, the word frequency information is updated if the confirmed homophone is not a compound word, and the compound word frequency information is updated if it is a compound word. It is.

A second feature is that when updating the frequency information of the compound word, the frequency information of the compound word is updated, and at the same time, the frequency information of each word constituting the compound word is also updated.

(Operation) According to the above configuration, compound words are learned over a long period of time in the same way as words, and erroneous conversion can be prevented when converting compound words into kana-kanji.

(Embodiment) Hereinafter, a document creation device according to an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing its configuration. In FIG. 1, a keyboard 11 outputs input information necessary for document creation etc. to an input/output control section 12. This keyboard 11 includes, in addition to "letter" keys, "numeric" keys, "cursor" keys, etc., a "conversion/next candidate" key for instructing kana-kanji conversion and homophone selection; A "selection/execution" key 11b for giving instructions is provided. The input/output control section 12 controls input/output of the keyboard 11, display section 13, etc., and here,
Controls reading input, pressure of kana-kanji conversion results, etc.

The display unit 13 is composed of, for example, a liquid crystal display, and displays the input reading and the kana-kanji conversion results for the reading.

The CPU 14 is in charge of controlling the entire device. The kana-kanji conversion control section 15 controls the kana-kanji conversion processing section 16 and the long-term learning processing section 17 according to the kana-kanji conversion instruction. The kana-kanji conversion processing unit 16 refers to the dictionary 18, the standard long-term learning dictionary 19, and the compound word long-term learning dictionary 20 to perform kana-kanji conversion. The long-term learning processing unit 17 updates the frequency information of the standard long-term learning dictionary 19 according to the appearance frequency of words, and updates the frequency information of the compound word long-term learning dictionary 20 according to the appearance frequency of compound words.

Here, each dictionary 18, 19, and 20 will be explained.

The dictionary 18 stores in advance a plurality of words with the same reading and frequency information of each word. The frequency information of words registered in this dictionary 18 is set in advance, for example, at the time of shipment of the device.

On the other hand, the standard long-term learning dictionary 19 is a dictionary of words used in the long-term learning mode, and is provided in one-to-one correspondence with the dictionary 18, and contains a plurality of words for the same reading and frequency information of each word. remembered. However, the frequency information of words registered in this standard long-term learning dictionary 19 is updated by the long-term learning processing section 17. FIG. 2 shows the structure of the standard long-term learning dictionary 19. Standard long-term learning dictionary 1
Frequency information (long-term information) of each word registered in 9.
Depending on the usage frequency of the word, for example, "+1" is added to the initial value.
” will be updated.

Further, the compound word long-term learning dictionary 20 is a compound word dictionary used during long-term learning, and stores a plurality of compound words with the same reading and frequency information of each compound word. Frequency information of compound words registered in this compound word long-term learning dictionary 20 is also updated by the long-term learning processing section 17 similarly to the standard long-term learning dictionary 19. FIG. 3 shows the structure of the compound word long-term learning dictionary 2o. Each compound word frequency information (long-term information) registered in the compound word long-term learning dictionary 20 is changed from the initial value to, for example, "+1j" according to the usage frequency of the compound word.
It will be updated one by one.

Note that a compound word is a word that combines two or more words to form a word with a new meaning.

Next, the operation of this embodiment will be explained with reference to the flowcharts shown in FIGS. 4 and 5.

First, as shown in FIG. 4, when the reading is input through the keyboard 11 and the "conversion/next candidate" key 11a of the keyboard ti is operated (step AI), the kana-kanji conversion processing section 16 in the kana-kanji conversion control section 15 Through this process, kana-kanji conversion is performed (step A2). The kana-kanji conversion processing unit 16 first determines whether the input reading information is a compound word based on a predetermined rule (step A3). Specifically, the dictionary 18 and the standard long-term learning dictionary 19 are searched, and if the word cannot be converted into a desired series of words in one conversion operation, it is determined that the word is a compound word.

As a result, in the case of a compound word, the kana-kanji conversion processing unit 16
refers to the compound word long-term learning dictionary 20 to check whether the compound word corresponding to the pronunciation has already been long-term learned (registered), and if it has been long-term learned, a homophone candidate is determined based on the frequency information in the same dictionary 20. Determine the output order of "Convert/Next candidate"
In response to the operation of the key 118, homophone candidates are output to the display unit 13 (steps A4 to A6).

On the other hand, if it is determined in step S3 that the word is not a compound word, or if it is determined in step S4 that long-term learning has not been performed even if it is a compound word, the kana-kanji conversion processing unit 16 converts the standard long-term learning dictionary 19 into a compound word. In reference to,
Check whether each word constituting the word or compound word (compound word that has not been learned for a long time) has already been learned for a long time. If it has been learned for a long time, homophone candidates are determined based on the frequency information in the dictionary 19. The output order of the homophone candidates is determined, and the homophone candidates are displayed on the display unit 1 in response to the operation of the "conversion/next candidate" key 11a.
3 (steps A7, A8, A6).

Further, in step S7, if it is determined that the word has not been learned for a long time, the kana-kanji conversion processing unit 16
The output order of homophone candidates is determined based on the frequency information (fixed information set in advance) in step 8, and the "conversion/next candidate"
In response to the operation of the key Lla, homophone candidates are output to the display unit 13 (steps A9, A6).

In this way, the result of the kana-kanji conversion is displayed on the display section 13. At this time, if there are multiple words for the reading, that is, if there are multiple homophone candidates, the kana-kanji conversion processing unit 16 first selects the most frequent word as the first candidate and displays it on the display unit 13 based on the output order determined above. Output to. The display unit 13 displays this word with a symbol indicating an undefined state, such as overline, for example.

Here, if the displayed word is not the word you want, click
2. Operate the "Convert/Next Candidate" key 11a. By operating this "conversion/next candidate" key 11a, the kana-kanji conversion processing unit 1
6 outputs the next candidate word to the display unit 13. As a result, the next candidate word is displayed with an overline as described above. On the other hand, if the displayed word is the desired word, the "select/execute" key 11b is operated. This “selection/
By operating the "Execute" key 11b, the word is confirmed and the overline attached to the word disappears. Note that the same applies to compound words.

Next, the operation after selecting and confirming a word or compound word as described above will be explained.

As shown in FIG. 5, select/execute j key 11. When the homophone of a word or a compound word is determined by the operation b (step B1), the long-term learning processing section 17 in the kanji conversion control section 15 determines whether the determined homophone is a compound word or not. Judgment is made (step B2).

In the case of a compound word, the long-term learning processing unit 17 registers the compound word in the compound word long-term learning dictionary 20 (steps 83 to B5).
. Registering a compound word in the long-term learning dictionary 20 means counting (updating) the number of times a compound word has been used as a long-term frequency. That is, as shown in FIG.
When the compound word "sudden start" consisting of "sudden start" and "start" is determined during homophone conversion, if this compound word "sudden start" has already been used four times, its frequency information (long-term information) to "5".

In this case, due to the capacity of the compound word long-term learning dictionary 2D,
When registering a compound word, the long-term learning processing unit 17 uses the same dictionary 20.
Check whether the number of registered words exceeds a predetermined number (Step B3)
, if the predetermined number is exceeded, the infrequent compound words are deleted from the dictionary 20 (step B4). This makes it possible to prevent the compound word long-term learning dictionary 20 from overflowing. Note that the predetermined number is arbitrarily set in the long-term learning processing section 17 in advance.

On the other hand, in step B2, if it is determined that the word is not a compound word, that is, if it is determined that it is a word, the long-term learning processing section 17 registers the word in the standard long-term learning dictionary 19. Registering in the standard long-term learning dictionary 19 means, as in the case of the compound word long-term learning dictionary 20, counting (updating) the number of times a word has been used as a long-term frequency.

In this way, words and compounds are individually studied over a long period of time. Thereafter, when converting a word into kana-kanji, the words (homonyms) corresponding to the reading will be displayed in order of frequency of use based on the frequency information of the words registered in the standard long-term learning dictionary 19. Similarly, when converting compound words into kana-kanji, compound words (homonyms) for readings will be displayed in order of frequency of use based on the frequency information of compound words registered in the compound word long-term learning dictionary 20. .

I will explain in detail. For example, when you input "Kyu Hasshin", it is first determined whether the input reading information is a compound word or not. In this case, as described above, it is determined whether the word is a compound word by searching the dictionary 18 and the standard long-term learning dictionary 19. If the word is determined to be a compound word, it is determined whether the word has already been studied for a long period of time. At the beginning, still
It has not been learned as a compound word over a long period of time. Therefore, in such a case, homophone candidates for "Kyuharishin" will be output at the word level. In other words, “
"Kyu" and "Hasshin" are converted into kana-kanji, for example, "
"Sudden call", "Sudden start", "Sudden rash", etc. are output in this order.

In this case, the output order of homophone candidates differs depending on whether each word constituting the compound word has been studied over a long period of time. That is, if each word has been studied over a long period of time, the output order of homophone candidates is determined based on the frequency information (long-term information) of the standard long-term learning dictionary 19, and the homophone candidates are output in that order. On the other hand, if each word has not been studied for a long time, Dictionary 1
The output order of homophone candidates is determined based on the frequency information (fixed information) of No. 8, and the homophone candidates are output in that order.

Next, it is assumed that "sudden start" is selected and confirmed from among the homophone candidates. Since "sudden start" is a compound word, it is registered in the compound word long-term learning dictionary 20 as shown in FIG. 3, for example. After that, for example, if you enter ``Kyuushakaiko'', it will be checked whether it has been learned as a compound word over a long period of time. You can check whether or not there is long-term learning at the word level.

Here, "kyu" is already used in "sudden start" mentioned above, but it is not used alone, so
As shown in Figure 2, there is no long-term learning. Also, there is no long-term study for "1 Society". Therefore, homophone candidates for "kyu" and "shiyakai" are output in a prescribed frequency order (fixed frequency order in the dictionary 18). This results in
For example, if the homophone candidates for each of the words "old society" and "society" are high, "old society" will be output as the first candidate, and it is possible to prevent erroneous conversion of "society".

In the above embodiment, when registering a compound word, only the compound word is registered in the compound word long-term learning dictionary 20, but for example, at the same time as registering the compound word in the same dictionary 20, the compound word is Each word may be registered in the standard long-term learning dictionary 19 at a lower ratio than usual. The operation when configured in this way is shown in Figures 6 to 8.
This will be explained with reference to the figures.

That is, as shown in FIG.
．． When a homophone of a word or a compound word is determined by the operation in step 1b (step C1), the long-term learning processing section 17 in the kanji conversion control section 15 determines whether the determined homophone is a compound word or not. (Step C2). In the case of a compound word, the long-term learning processing unit 17 registers the compound word in the compound word long-term learning dictionary 20, and at the same time registers each word constituting the compound word in the standard long-term learning dictionary 19 (steps 03-
C6).

In this case, as in the above embodiment (FIG. 5), due to the capacity of the compound word long-term learning dictionary 20, the long-term learning processing unit 17 determines whether the number of registered words in the compound word 20 exceeds a predetermined number when registering the compound word. (Step C3) If the predetermined number is exceeded, the infrequent compound word is deleted from the dictionary 20 (Step C4).

In this way, for example, as shown in FIGS. 7 and 8, when the compound word "sudden start" is registered in the compound word long-term learning dictionary 20, the words "sudden" and "start" are registered. will also be registered in the standard long-term learning dictionary 19. In addition,
Updating the word frequency in this case is different from the case of long-term learning of words alone, so it needs to be set lower than usual. As a result, you can simultaneously study a compound word and each word that makes up the compound word over a long period of time, and from now on, when you input ``kiyutenbo,'' for example, the ``kyu'' you are learning will be given priority. It is possible to output and display "rapid tempo" as the first candidate.

Further, in the above embodiment, the long-term learning dictionary is divided into standard and compound words, but the present invention is not limited to this. For example, as shown in FIG. 9, the long-term learning dictionary may be constructed as two long-term learning dictionaries. is also possible. In this case, a compound word flag A may be provided for each word, and when the word is used as a compound word, the compound word flag A may be set for that word.

[Effects of the Invention] As described above, according to the present invention, compound words can be studied over a long period of time in the same way as words. Therefore, when converting a compound word into kana-kanji, a desired word can be quickly obtained and erroneous conversion can be prevented. This greatly improves document creation efficiency.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a document creation device according to an embodiment of the present invention, and FIG. 2 shows a standard long-term study frame of the same embodiment.
Figure 3 is a diagram showing the configuration of the compound word long-term learning dictionary of the same embodiment, Figure 4 is a flowchart for explaining the operation of the kana-kanji conversion process of the same embodiment, and Figure 5 is a diagram showing the structure of the compound word long-term learning dictionary of the same embodiment. FIG. 6 is a flowchart for explaining the operation of the long-term learning process according to another embodiment of the present invention, and the structure of the first compound word long-term learning dictionary. FIG. 9 is a diagram showing the structure of a long-term learning dictionary for standard and compound words according to another embodiment of the present invention. 11...Keyboard, lla..."Conversion/Next candidate"
Key 11b..."selection/execution" key I2...input/output control section, 13...display section, 14...CPU 515
...Kana-Kanji conversion control, 16...Kana-Kanji conversion processing unit, 17...Long-term learning processing unit, 18...Dictionary, 19
...Standard long-term learning dictionary, 20...Compound word long-term learning dictionary. Applicant's agent Patent attorney Takehiko Suzue

Claims (3)

[Claims] (1) A first storage means that stores a plurality of words for the same reading and frequency information of each word; and a second storage means that stores a plurality of compound words for the same reading and frequency information of each compound word. , a first compound word determining means for determining whether or not the input reading information is a compound word; and a first compound word determining means for determining whether or not the input reading information is a compound word; Candidate extraction for extracting homophone candidates from the first storage means, and extracting homophone candidates of the compound word corresponding to the input reading information from the second storage means when it is determined that the compound word is the compound word. a homophone selection confirmation means for selecting a desired homophone from among the homophone candidates extracted by the candidate extraction means and confirming the selected homophone; a second compound word determining means for determining whether a homophone is a compound word; and a second compound word determining means for updating word frequency information when the second compound word determining means determines that the word is not a compound word; What is claimed is: 1. A document creation device having a learning function, comprising: an updating means for updating frequency information of a compound word when it is determined that there is a compound word. (2) The updating means updates the frequency information of the compound word when it is determined to be a compound word by the second compound word determining means, and at the same time updates the frequency information of each word constituting the compound word. A document creation device with a learning function according to claim 1. (3) In a document creation device equipped with a learning function, a plurality of words corresponding to the same pronunciation and frequency information of each word are stored in a first storage area of the storage means, and a plurality of compound words corresponding to the same pronunciation and frequency information of each of these words are stored in a first storage area of the storage means. Frequency information of compound words is stored in a second storage area of the same storage means, and it is determined whether the input reading information is a compound word or not.As a result, if it is determined that the input reading information is not a compound word, the above input reading information is stored. A homophone candidate for the word corresponding to the input reading information is extracted from the first storage area, and when it is determined that the word is a compound word, the homophone candidate for the compound word corresponding to the input reading information is extracted from the second storage area. When a desired homophone is selected and confirmed from the extracted homophone candidates, it is determined whether or not this homophone is a compound word. A learning method characterized in that frequency information of a word is updated when it is determined that the word is a compound word, and frequency information of the compound word is updated when it is determined that the word is a compound word.