JPH0683807A - Document preparing device - Google Patents

Abstract

PURPOSE:To improve operability by reducing the generation of clause segmenting error and improving the probability of conversion to a sentence desirable for a user by learning clause segmenting information. CONSTITUTION:When a character string is inputted from an input part 11 and KANA (Japanese syllabary)/KANJI (Chinese character) conversion is instructed, a KANA/KANJI conversion control part 13 performs KANA/KANJI conversion processing to the inputted character string, stores the result in a candidate memory 17 and displays the preferential converted system at a display part 23. When the segmentation of the clause is instructed at such a time, the segmentation control of the clause is performed under the control of a clause segmentation control part 14. A clause learning control part 18 refers to a clause learning dictionary 19, prepares the converted system according to the segmented clause and outputs the preferential system. When the converted system is fixed after the clause is segmented, the clause learning dictionary 19 stores clause cut information composed of one part, part of speech and connection information, etc., of reading on the front side rather than the designated segmentation and the reading of the following clause as learning information.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document preparation device having a phrase recutting function.

[0002]

2. Description of the Related Art In a document creation apparatus, when a character string including a plurality of clauses is converted into Kana-Kanji characters at once, for example, in the case of a character string "Kazoku hanare kurashi", (a) A possible example is the phrase cutting "Kutaku / Hanare / Living" (b) "Kakuzoku / Harelet / Living".

According to information such as word information stored in a dictionary as an internal dictionary, information about auxiliary words such as particles, and information on usage of adjectives, the document creation apparatus uses, for example, the longest matching method (a). A conversion sequence is created by performing any clause break of (b). Therefore, for example, even if the character string is input in the expectation of the phrase breakage of (b), the disconnection of (a) may be performed.

As a countermeasure against this, the conventional document creating apparatus has a function of re-cutting the clauses so that the Kana-Kanji conversion desired by the user is performed when the correct Kana-Kanji conversion is not performed due to the difference in the position of the clause break. I have it.

[0005] For example, when a character string "Kazoku is far away" is entered and converted to "Living with a family," it becomes a sentence "Living with a family." ,
Specify the first phrase, "kazokuto," to instruct the phrase to be recut. As a result, the "kazokuto" part and the "haarete kurusu" part are converted into kana-kanji characters again, and the sentence becomes "living away from family".

[0006]

In the conventional document creating apparatus, when the phrase is re-cut, the first re-cutting, that is, the previous phrase is regarded as one word and the dictionary is searched. Therefore, if you perform a phrase recut including a compound word, the corresponding kanji cannot be searched,
Correct Kana-Kanji conversion is not performed.

For example, when a character string "Emoji no Nyuryoku" is input and converted into "Emoji input", this can be changed to "Emoji input" by the first clause. Specify a certain "e" part to instruct the phrase recutting. As a result, the “e” part and the “moji no nikyuryuku” part are converted into kana-kanji characters again and can be used as “input of pictographs”. However, if the phrase is shuffled by specifying "Emoji no", it will not be converted correctly. This is because each of "E" and "Momoji" is a bunsetsu, and in "Emoji no", there are two bunsetsu, so the part before the recut position, that is, "Emoji no", is not converted to Kanji. Is.

Further, since the prefix and the suffix are also regarded as one clause, the following example also occurs. For example, if you enter a character string such as "Kousei no Uhin no Kaihatsu" and it is converted to "Development of component delivery", if you want to change it to "Development of high performance product", specify with "Kou" And instruct to re-cut the phrase. As a result, the "Kou" and "Seino Uhin no Kaihatsu" parts are converted into kana-kanji characters again, and "high-performance product development" can be achieved. However, if the phrase is re-chosen by specifying "Koseinou", two phrases including a compound word are designated, and the "Kosei-inou" part designated by recutting is not converted into Kanji.

As described above, in the conventional document creating apparatus, the first one phrase must be designated as a unit to instruct the phrase reshuffling, and the phrase including the compound word must be designated to instruct the phrase reshuffling. In this case, there was a problem that the front part specified to be re-cut was not converted to kanji and kana-kanji conversion was not performed reliably.

The present invention has been made in order to solve the above problem, and even if a phrase is re-cut including a compound word, the learning can surely perform the kana-kanji conversion that the user desires. An object is to provide a producing device.

[0011]

A document creating apparatus according to the present invention is a kana-kanji converting means for creating a conversion sequence of kanji-mixed sentences by cutting a reading character string including a plurality of phrases into severable bunsetsu candidates. , Kana-Kanji conversion by referring to this phrase learning dictionary and the phrase learning dictionary that stores the phrase reading information that contains the compound reading and the following part of the reading of the following phrase, part of speech, connection information, etc. Means for selecting and outputting a sequence to be prioritized from the conversion sequences converted by the means, means for designating a break in the conversion sequence output by this means and instructing the phrase re-sizing, and this means A phrase re-slicing means for re-slicing a phrase by referring to the phrase learning dictionary according to a specified break, and a re-segmentation by this means, and a reading before the specified break,
It is characterized by further comprising learning means for updating the stored contents of the phrase learning dictionary according to the phrase cutting information including part of the subsequent phrase reading, part of speech, connection information, and the like.

[0012]

When a character string is input from the input unit and a kana-kanji conversion instruction is given, the kana-kanji conversion means cuts the input character string into severable phrase candidates and creates a conversion sequence of kanji-mixed sentences. Then, a sequence to be prioritized is selected from the converted sequences by referring to the phrase learning dictionary and displayed on the display unit.
At this time, if the phrase is accidentally cut and the desired conversion sentence cannot be obtained, a break is specified and the phrase is recut.
When this bunsetsu reshuffle is instructed, the bunsetsu learning dictionary is referenced, and a new conversion sequence including a compound word is created before the specified break. Then, when this new conversion sequence is confirmed, the phrase learning is performed based on the reading before the specified break, a part of the reading of the succeeding phrase that follows the break, and the phrase cutting information including the part of speech and connection information. Update the memory contents of the dictionary.

When a new reading character string is input, the bunsetsu learning dictionary is referenced to cut the bunsetsu, so that the occurrence of bunsetsu error is extremely reduced, and the sentence desired by the user is reduced. The conversion probability can be increased, and the operability can be significantly improved.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the arrangement of a document creating apparatus according to an embodiment of the present invention. In FIG. 1, an input unit 11 is a keyboard for inputting various kinds of information necessary for creating a document. The input unit 11 is provided with various function keys such as a "conversion / next candidate" key, a "select / execute" key, and a "phrase recut" key in addition to the "character" key, the "numerical value" key, and the like. There is. Various input information from the input unit 11 is input to the Kana-Kanji conversion control unit 13 via the input control unit 12. A phrase recutting control unit 14 is connected to the input control unit 12 and the Kana-Kanji conversion control unit 13. These input control unit 12, Kanji conversion control unit 1
The RAM 15 is connected to the 3 and the phrase recut control section 14. The RAM 15 is provided with various areas such as an area 15a for temporarily storing a phrase recutting position and a document information area 15b for storing created document information.

The kana-kanji conversion control unit 13 refers to the kana-kanji conversion dictionary 16 to perform kana-kanji conversion, and stores the created sentence in the document information area 15b of the RAM 15.
The kana-kanji conversion dictionary 16 stores reading, a word (converted word) corresponding to the reading, grammatical information of the word, and semantic information indicating, for example, what the word means. The bunsetsu recutting control unit 14 performs bunsetsu recutting control when a bunsetsu recutting instruction is given.

Further, the Kana-Kanji conversion control unit 13 includes:
The candidate memory 17 and the phrase learning control unit 18 are connected. The candidate memory 17 stores a phrase candidate for the input character string and a conversion sequence in which the phrase candidates are connected in descending order of relevance. The bunsetsu learning control unit 18 refers to the bunsetsu learning dictionary 19 to determine a bunsetsu to be prioritized for the input character string, creates a plurality of conversion sequences, and registers the bunsetsu cutting information to be learned in the bunsetsu learning dictionary 19. To do.

The kana-kanji conversion control unit 13 outputs the document information stored in the document information area 15b to the display control unit 21. The display control unit 21 draws the document information and the like sent from the Kana-Kanji conversion control unit 13 in the display memory 22, and displays the image data on the display unit 23 such as a backlit LCD panel.

As shown in FIG. 2, the bunsetsu learning dictionary 19 reads the bunsetsu including the compound word on the front side as bunsetsu cutting information indicating a break between the bunsetsu containing the compound word on the front side and the bunsetsu word on the back side.
A part of the reading of the later clause, for example, reading of the stem, part of speech, connection information, and other information is registered. In FIG. 2, the case where the reading on the front side is “Emoji” and the reading of the subsequent clause is “Nyuryoku” is shown as an example.

As described above, the reading on the front side is the reading of a phrase containing a compound word. This allows one aggregated portion to be designated as a target for phrase cutting, which makes it less likely that a cutting error will occur and that the phrase will be This is to ensure that the kana-kanji conversion will be performed even if the character string before the break is a bunsetsu containing a compound word when recut.

[0021] Further, the reason why the reading of the front side and the succeeding phrase are registered is that the cutting of the phrase is generally determined by the entanglement with the succeeding phrase, and the registration of only the reading of the front side is performed. , This is because there is a problem that the subsequent clauses become uniform in any case.

Further, a part of the reading of the subsequent phrase, for example, only the stem of the word is memorized, because the particle following the stem or the inflection ending changes variously, and if the entire reading of the succeeding phrase is remembered. This is because the learning is extremely narrow. Furthermore, the part of speech of the stem of the later bunsetsu is remembered because it may have a completely different meaning. Next, the operation of the above embodiment will be described. As shown in the flow chart of FIG. 3, a character string to be converted from the input unit 11, for example, “Kousei no Uhin no Kaihatsu”

Input the character string (step A1),
When kana-kanji conversion is instructed by the "conversion / next candidate" key (step A2), the input information is input to the kana-kanji conversion control unit 13 via the input control unit 12. As a result, the kana-kanji conversion control unit 13 separates the clauses,
Performs Kana-Kanji conversion processing. That is, the kana-kanji conversion control unit 13 refers to the kana-kanji conversion dictionary 16 and performs a grammatical analysis such as to which part of each bunsetsu is the stem and where is the attached word or the inflection ending, and all bunsetsu candidates obtained as a result are analyzed. Store in candidate memory 17 (steps A3, A4
).

Here, the phrase learning control section 18 refers to the phrase learning dictionary 19 shown in FIG. 2 to judge which of the phrase candidates stored in the candidate memory 17 is highly relevant and should be prioritized (step A5, A6) and a conversion sequence are created (step A7).

Next, the phrase learning control unit 18 refers to the phrase learning dictionary 19 to determine which of the conversion sequences stored in the candidate memory 17 should be output preferentially (step A).
8), the highest priority one is output to the display control unit 21 as the conversion result, displayed on the display unit 23 and waits for key input such as phrase recutting (steps A9, A10).

In this state, it is judged whether or not there is an instruction to re-phrase the phrase (step A11). If there is no instruction to re-phrase the phrase and the confirm key is operated, the process is terminated. However, if the phrase recutting is instructed, the phrase recutting process is executed under the control of the phrase recutting control unit 14 based on the instruction. Now, for example, when the character string "Kousei no Hinhin Kaihatsu" is entered, the phrase "Kosei / Nouhinno / Kaihatsu" is cut and converted to "Development of component delivery". To do. For this conversion character, the user can "develop a high-performance product" by cutting the phrase "Kou / Seinou / Hinno / Kaihatsu".

If you want to convert the kana to kana characters,
Specify the character string "Kou" or "Kousei Nou" to instruct the phrase recutting. This phrase re-positioning is specified by, for example, the "function 1" key and the cursor key "←",
The position of the phrase break is specified by the combination operation of "→", and the execution of the phrase recut is instructed by the operation of the "select / execute" key. The position of the phrase recutting is indicated by, for example, inversion of the displayed character. The position information of the phrase re-cutting is stored in the phrase re-cutting position storage area 15 of the RAM 15.
It is temporarily stored in a.

When the phrase recutting is instructed, the phrase recut control unit 14 controls the phrase recut. That is, the kana-kanji conversion control unit 13 refers to the position information for phrase re-cutting and the kana-kanji conversion dictionary 16 (step A12) and performs grammatical analysis to create a phrase candidate (step A13). Here, the bunsetsu learning control unit 18 refers to the bunsetsu learning dictionary 19 to determine the bunsetsu to be preferentially output among the bunsetsu candidates stored in the candidate memory 17 (steps A14 and A15), and determines the conversion sequence. Create and validate only the series that is cut at the shuffle position (step A
16). In this case, the re-cut front character string is subjected to kana-kanji conversion processing even in reading including a compound word. For example, even if a character string containing a compound word such as "Kousei Nou" is shuffled, conversion characters such as "high performance" are validated, and multiple conversion series including "development of high performance products" are performed. Judge as a valid series.

Further, the phrase learning control unit 18 invalidates a sequence in which the reconverted front side conversion character has two or more phrases for the above conversion sequence (step A17). For example, when a character string of "Emoji no" is input, it is converted to the compound word "Emoji no", and conversion including two phrases such as "Emomo no No" is performed. The conversion accuracy is improved by invalidating the series such as "picture also character" in which the conversion character of 2 is two or more clauses.

Next, the phrase learning control unit 18 refers to the phrase learning dictionary 19 to determine the sequence to be prioritized among the conversion sequences stored in the candidate memory 17 (step A18), and the highest priority one, for example, ""Development of high-performance product" is output to the display control unit 21 as the conversion result and displayed on the display unit 23 (step A19). In this state, it is determined whether or not the confirmation operation has been performed (step A20).

When the result displayed on the display section 23 is not what the user desires, the cursor is moved to the target clause position and the "conversion / next candidate" key is operated.
By this key operation, the phrase candidate having the next highest priority is read from the candidate memory 17 and displayed on the display unit 23. Further, by designating “between homophones”, all the phrase candidates of the “homophone” stored in the candidate memory 17 are read out and displayed collectively in the editing area of the display unit 23.
A desired phrase candidate may be selected and designated from among them. When the recut operation is performed again, the process returns to step A12 via step A11, and the same operation as described above is repeatedly executed.

When the kana-kanji desired by the user is performed as described above, the converted sentence is confirmed by operating the "select / execute" key. Further, by performing this confirming operation, the content stored in the phrase re-positioning area storage area 15a of the RAM 15 is cleared, and new phrase cutting information associated with the phrase re-cutting is added to the phrase learning dictionary 19.
Is registered (step A21). That is, when the phrase is cut, the kana-kanji conversion dictionary 16 is searched for the character string, and grammatical checks such as word extraction, part-of-speech check, and connection relation check are performed. Therefore, the phrase cutting information is registered in the phrase learning dictionary 19. For example, in the process of re-phrasing a phrase, for the character string of "Koseinou Hinno Kaihatsu", (1) Cut the phrase with "U" of "Koseinou" (2) "Kou" and "Seino""U" is registered as a word in the dictionary. (3) Search the dictionary with "Hinno Kaihatsu". (4) "Hin": There is a noun word in the dictionary. (5) "No": Registered as a particle in the dictionary and connectable with "Kaihatsu" (noun). (6) "Kaihats": Registered as a noun in the dictionary.

The above-mentioned operation is carried out, and a part of the reading of the front side and the reading of the succeeding bunsetsu, such as the reading of the stem, the length, the part of speech, the connection information, and other information obtained in this process, It is registered in the phrase learning dictionary 19. The registered content of the phrase learning dictionary 19 is searched when a new reading character string is input, and when the phrase cutting information corresponding to the reading of this character string is registered, Phrase break is performed according to the phrase break information.
As a result, the Kana-to-Kanji conversion desired by the user will be performed.

In the above embodiment, step A in FIG.
In 17, the re-converted front side conversion character is valid only for the series of compound words, and the one consisting of multiple clauses is invalid, but without performing this process, the re-converted front side conversion character is Even if two or more clauses are included, it may be determined to be valid. In this case, kana-kanji conversion can be performed even if the re-read front reading includes a compound word or two or more phrases. Therefore, in this case, since the sequence including the compound word and the sequence including two or more phrases in the reconverted conversion character on the front side are created, even if the displayed conversion sentence is not the sentence desired by the user, the phrase It is possible to select and specify from the candidate sentences without performing re-cutting again.

[0035]

As described above in detail, according to the present invention, the reading of the front side including the compound word in the phrase learning dictionary and a part of the reading of the succeeding phrase, the part of speech, the connection information, and other information. The phrase break information consisting of 1 is registered, and the phrase break information is learned by re-cutting the phrase.
It is possible to target bunsetsu breaks to one grouped part, and even when instructing re-cuts of bunsetsu including multiple words, the occurrence of bunsetsu error is very small, and as a result, the sentence that the user wants The conversion probability of can be increased, and the operability can be significantly improved.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of a document creation device according to an embodiment of the present invention.

FIG. 2 is a diagram showing a configuration of a phrase learning dictionary in the embodiment.

FIG. 3 is a flowchart showing the operation of the embodiment.

[Explanation of symbols]

11 ... Input unit, 12 ... Input control unit, 13 ... Kana-Kanji conversion control unit, 14 ... Phrase recutting control unit, 15
... RAM, 16 ... Kana-Kanji conversion dictionary, 17 ... Candidate memory, 18 ... Phrase learning control unit, 19 ... Phrase learning dictionary, 21 ... Display control unit, 22 ... Display memory,
23 ... Display section.

Claims (1)

[Claims] 1. A kana-kanji conversion means for creating a conversion sequence of kanji-mixed sentences by cutting a reading character string including a plurality of phrases into severable bunsetsu candidates, and a front-side reading including a compound word and a subsequent reading. Refer to this phrase learning dictionary and the phrase learning dictionary that stores the phrase disconnection information consisting of part of the reading of the phrase, part of speech, connection information, etc., and refer to this phrase learning dictionary, and give priority to the conversion sequence converted by the above Kana-Kanji conversion means. A means for selecting and outputting a sequence, a means for designating a break in the conversion sequence output from this means and instructing the phrase reshuffling, and referring to the above phrase learning dictionary by the break specified by this means The phrase re-slicing means, and the phrase re-slicing by this means, the segmental part consisting of the part of the segmental part of speech, the part of the segment of the segmental part of the segmental part of the segmental part of the segmental part of the segmental part of the segmental part which is read before the break specified above A document creating apparatus comprising: learning means for updating the stored contents of the phrase learning dictionary according to disconnection information.