JPS6132169A - Word extracting system - Google Patents

Abstract

PURPOSE:To extract a word with suffix registered in a word dictionary while decreasing the reduction in retrieval processing speed by generating three character strings only when the 3rd Kanji (Chinese syllabary) sound has a possibility of a suffix. CONSTITUTION:A sectioning section 3 using a Kanji sound applies sectioning by means of the Kanji sound to an input Kana (Japanese syllabary) character string transmitted from an input character string storage setion 2 at first. An attribute discriminating section 5 of the input character string discriminates whether there is any possibility of the 3rd minimum unit to be a suffix or not depending on the arrangement of the Kanji sound ans the Kana of the input character string sectioned by using the Kanji sound and the Kana as the minimum unit. Only when there is a possibility, the retrieved character string is generated based on the three minimum units and in other cases, the retrieved character string is formed based on the two minimum units as a conventionally system.

Description

DETAILED DESCRIPTION OF THE INVENTION Technical Field The present invention relates to a kana-kanji conversion processing device, and more particularly to a word extraction method in a kana-kanji conversion processing device that can be applied to a Japanese document creation device, a computer system, etc.

In the conventional kana-kanji conversion processing device, a word dictionary for kana-kanji conversion is provided in order to convert sentences input in phonetic characters (hiragana, katakana, Roman letters) into appropriate sentences containing kanji and kana. . This search of the word dictionary is performed by cutting out a word from the input kana character string as a character string to be searched, and then matching the character string to be searched with the heading character string in the word dictionary. However, since Japanese has a complex grammar and many homonyms, a dictionary search will extract multiple candidate words.

Conventionally, the following processing is performed to select one of the plurality of candidate words and use it as a conversion result. In other words, for each extracted candidate word, the possibility of connection with the previous converted word (conversion result) is determined, and the connectable candidate words are determined using parameters such as reading length, frequency of appearance, connection weight, etc. The candidate word with the highest evaluation is output as the conversion result.

By the way, conventionally, in order to facilitate word dictionary searches and reduce misunderstandings, input character strings have been preprocessed using kanji sounds.

Kanji sounds, when written in kana, include ■1-character kanji sounds whose length is one character, ■2-character kanji sounds whose length is two characters, and ■three-character kanji sounds whose length is three characters. For example, ■1 character kanji sound... is the most Kananian "A,", I "
■Two-character kanji sound...Ai "Love, Greeting, Sadness...", Ak "
Evil, grip...'', etc.■ Three kanji sounds...Shu ``Collect, collect, week, shu, end, xi,
``xiu, zhou, shu...'' and sho ``xiang, xiao, shu, katsu, xiao, shang, sho, erasing, zheng...'' and so on.

By the way, in the above-mentioned 2-character kanji sounds and 3-character kanji sounds, the kana that occupy the second and third characters are limited to the following 18 types.

“I, tsu, ki, ri, chi, tsu, ya, yu, yo, yuu, yo.
Yaku, yuku, yo, yutsu, yun, tsu, n' But 18
Not all types of kana form a kanji sound for the first kana. For example, if the first character is A.

Ai...Kanji sound (see example above) Aku...not a kanji sound Aki...not a kanji sound Ak...kanji sound (see example above) Achi...not a kanji sound become that way.

By dividing the input character string by the above two or more kanji sounds and creating a search string using these as units, it is possible to avoid misinterpreting what is originally part of the kanji reading as a case particle, etc. . Furthermore, the above Kanji sounds are never used alone, but are always used in combination with other Kanji sounds. Therefore, as a result of matching an input string of three or more characters with these kanji sounds, if the first character is matched but the second and third characters are not, the first character is not a kanji sound but an attached character. It can be inferred that it is likely to be a kana such as a word. For this reason, it is not necessary to store all kanji sounds, including single-character kanji sounds, in the kanji sound device for kanji sound searches, as shown in Figures 3 (,), (b), and (c).
It is sufficient to store two or more kanji sounds as shown in .

Conventionally, the following preprocessing has been performed using kanji sound instruments as shown in FIGS. 3(a), 3(b), and 3(c).

For example, the kanji sound system is accessed based on the input kana character string ``Bunsho no sakusei wa hijyo ni yoidei.'' and the following separation is performed using kanji sounds and kana as the minimum units.

"jin/sho/no/saku/sei/ga/hi/jiyou/ni/you/i/de/a/ru/." However, katakana indicates kanji sounds, and hiragana indicates kana.

After the above delimiting process, attributes are added to the human character string as follows, depending on how the kanji sounds and kana are arranged.

(Kanji sound) + (Kanji sound) 10~ ・・・・・・TYPEI
(Kanji sound) + (kana) 10~ ・・・・・・TYPE2
(Kana) + (Kanji sound) 10~ ・・・・・・TYPE3
(Kana) + (Kana) 10~ TYPE 4
Bun+shou+~'' is (kanji sound) + (kanji sound) ten~, so the above example sentence becomes TYPEI.

Next, a searched character string is created in the following manner according to the attributes TYPEI to TYPE4 of the input character string.

For TYPEL...■(Kanji sound) + (Kanji sound)■(Kanji sound) For TYPE2...■(Kanji sound)+(Kana)
■(Kanji sound) For TYPE3...■(Kana) + (Kanji sound)
■(kana) For TYPE4...■(kana)■(kana)+
(Kana) ■ (Kana) + (Kana) + (Kana) In the above example sentence, since it is TYPEI, set the searched character string as follows.

[The set search string... Select candidate words.

Here, it is assumed that [sentence J is extracted as the optimal candidate word. In this case, the next character string to be analyzed is ``Nosakusei ga hijyo ni yoi desu.'', so this is again divided into kanji sounds and kana as the minimum units.

Y's /saku/sei/ga/hi/jiyou/ni/you/i/de/a/ru. "[no+sakuju~" is (kana) + (kanji sound) ten~, so the attribute of the above input character string is TYPE3. TYPE
When a searched character string is created according to the attributes of 3, the searched character string will be as follows.

[The search string to be set... After ■ no saku ■, the remaining input strings are separated using kanji sounds and kana as the minimum units, and the kanji sounds and kana are separated according to the arrangement of kanji sounds and kana. Create an appropriate search string according to the attributes of the input string.

The reason for creating a search string using two minimum units is that most of the words registered in the word dictionary are
This is to speed up the dictionary search, taking into account the fact that the minimum unit is smaller than 1.

However, the above method has the following drawbacks when a suffixed word exists in the input character string.

The input character string "Setsumeikaideha-・" will be explained as an example.

In this case, pre-processing using kanji sounds allows
The words are separated as follows: kai/de/ha~. "Set+meiju~" is (kanji sound) + (kanji sound) ten~, so it belongs to TYPEL.
■Setsumei■Setsu is set as the character string to be searched.

Therefore, even if "briefing session" is registered in the word dictionary, it will not be searched, and in order to obtain the kana-kanji conversion result "briefing session."

It is necessary to search (analyze) the word dictionary at least twice for "setsumei" and "kai". Furthermore, as a result of two rounds of analysis, the suffix ``kai'' is not necessarily selected, but ``kai'', ``kai'', ``kai''/``kai'', ``kaij, etc.'' are selected. The possibility of erroneous analysis also increases.

Purpose The purpose of the present invention is to eliminate the above-mentioned drawbacks of the prior art, to suppress a decrease in the word dictionary search speed when extracting words in a kana-kanji conversion processing device, and to extract words with suffixes registered in the word dictionary. The objective is to extract words reliably and effectively.

Configuration In order to achieve the above object, the word extraction method according to the present invention is provided in a kana-kanji conversion processing device having a means for dividing an input character string using kanji sounds and kana as the minimum units. a first means for determining the arrangement of kanji sounds and kana; a second means for determining whether a suffix having the same reading as the kanji sounds of a character string to be analyzed exists; and at least a character string to be analyzed. If the first or second minimum unit of is a kanji sound, and the third is a kanji sound with the same reading as the suffix, a third means of setting the searched character string by the three minimum units. It is distinctive in that it has been established.

Note that since the input character strings are sequentially analyzed, the input character string that should be analyzed for the time being naturally changes one after another.

In this specification, an input character string to be analyzed for the time being is named an analysis target character string.

Hereinafter, the configuration of the present invention will be explained in detail using an example.

FIG. 1 is a block diagram of a kana-kanji conversion processing device to which a word extraction method according to an embodiment of the present invention is applied.

In Fig. 1, ■ is a kana character input section for inputting a kana character string corresponding to the Japanese sentence to be created, 2 is an input character string storage section for temporarily storing the input kana character string, and 3 is an input character string storage section for temporarily storing the input kana character string. 4 is a kanji sound separation section that divides the input character string using kanji sounds and kana as the minimum unit, and 4 is a kanji sound similar to that shown in Figures 3 (a), (b), and (C). 5 is an input character string attribute determination unit that determines the attribute of a kana character string separated by kanji sounds based on the arrangement of kanji sounds and kana; 6 is an attribute judgment unit for determining the attributes of a kana character string separated by kanji sounds; The kanji sound attribute determination unit 7, which determines the attribute based on whether or not a suffix with the same reading exists, is obtained from the input character string attribute determination unit 5 and the kanji sound attribute determination unit 6. 9 is a word dictionary; 9 is a word dictionary; 9 is a word dictionary;
8 is a dictionary search unit that searches the word dictionary 9 based on the character string to be searched; 10 is a candidate word storage unit that stores candidate words obtained from the dictionary search unit 8; and 11 is a unit that evaluates the candidate words and selects the most appropriate word. A candidate word evaluation unit that selects a candidate word, 12 an optimal candidate word storage unit that stores the optimal candidate word selected by the candidate word evaluation unit 11, and 13 a display unit that displays the optimal candidate word as a kana-kanji conversion result. It is.

The attribute determining unit 5 based on kanji sounds determines which of the following two attributes each kanji sound constituting the character string to be analyzed belongs to.

Kanji sound attribute A: There is a suffix with the same reading as the kanji sound.

Examples: Deki (target), Kai (kai), etc. Kanji sound attribute B: There is no suffix that has the same reading as the kanji sound.

Examples: Ak, Setsu, Iki, etc. FIG. 2 is a flowchart showing the operation of a word extraction method according to an embodiment of the present invention.

First, the delimiter 3 based on kanji sounds is input to the input character string storage 2.
The input kana character string sent from is divided by kanji sounds using the kanji sound device 4 (201). The input character string attribute determination unit 5 determines whether the attribute of the input character string (character string to be analyzed) separated by Kanji sounds and kana as minimum units is TYPEI, TYPE2, or TYPE3, based on the arrangement of Kanji sounds and kana. (202). T.Y.
In cases other than PEI, TYPE2, and TYPE3, set the searched character string using the conventional method (203) TYPE
In the case of TYPE 1, TYPE 2, or TYPE 3, a search target character string setting process unique to this embodiment described below is executed. The reason is TYPEI...(Kanji sound) + (Kanji sound) 10~TYPE2...(Kanji sound) + (Kana) 10~T
This is because in the case of YPE3... (kana) + (kanji sound) 10~, the third minimum unit may be a suffix.

In this embodiment, since attributes A and B of the kanji sounds are used, after making the determination in step 202, it is first determined whether or not the character string to be analyzed is TYPE1.
If so, it is determined whether the third minimum unit is a kanji sound (2o4.2°5). When the third minimum unit is a kanji sound, it is further determined whether the kanji sound has attribute A, and if it is attribute A, a searched character string is created using the three minimum units (205.206. 2o7). The character strings to be searched are three: ■ (Kanji sound) + (Kanji sound) + (Kanji sound of attribute A) ■ (Kanji sound) + (Kanji sound) ■ (Kanji sound).

For example, in the case of the input character string "Setsumei Kaideha~", the kanji sound delimiter is "Set/Mei/Kai/De/Ha~", and the searched character string is ■Setsumeikai■Setsumei■Setsuyo Set.

Among the above three searched character strings, [nosetumeikai] has the longest reading length and is highly rated, so if "briefing session" is registered in the word dictionary ” is selected. Also, in the case of longest term - teaching method, "explanation session" is naturally selected.

The reason for determining the possibility that the third kanji sound is a suffix (attribute A) is that even if the first and second minimum units are suffixes, they cannot be extracted using the conventional method. Because there is.

If the third minimum unit of TYPEI is not a Kanji sound, or if the attribute is A even if it is a Kanji sound, the searched character string is set using the conventional method (■).

As mentioned above, only when there is a possibility that the third kanji sound is a suffix, a search string is created based on three minimum units, and in other cases, two By creating the searched character string based on the minimum unit of , it is possible to suppress the rate of decrease in the search processing speed for the word dictionary 9.

If it is not TYPEI, determine whether it is TYPE2 or not,
If it is TYPEI, it is determined whether the third minimum unit is a kanji sound (204.208.209). If the third minimum unit is a kanji sound, it is further determined whether the kanji sound has attribute A, and if it is attribute A, a search character string is created using the three minimum units (209, 210, 21
1).

In this case, the character strings to be searched are three: (1) (kanji sound) + (kana) 10 (character sound of attribute A) (2) (kanji sound) + (kana) + (kanji sound) (2) (kanji sound).

When the third minimum unit of TYPE 2 is not a Kanji sound, or even if it is a Kanji sound, the attribute is A, the searched character string is set using the conventional method (■).

If it is determined in step 208 that it is not TYPE 2, the character string to be analyzed can be identified as TYPE 3, so it is determined whether the third minimum unit is a kanji sound (208.212). If it is a kanji sound, further check its attributes, and if the attribute is A, create a search string using three minimum units (212.213.214)
. The character strings to be searched in this case are three (kana) + (kanji sound) + (kanji sound of attribute A) - (kana) + (kanji sound) - (kana).

Finally, the word dictionary 9 is searched using the obtained searched character string (215).

In this way, by creating a search string using three minimum units only when the third minimum unit is a Kanji sound and there is a possibility that the Kanji sound is a suffix,
It is possible to extract words with suffixes registered in a word dictionary while suppressing a decrease in search processing speed.

Effects As explained above, according to the word oil path method of the present invention,
Extracting words using a kana-kanji conversion processing device; while suppressing a decrease in word dictionary search speed.

Words with suffixes registered in a word dictionary can be extracted reliably and effectively.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a kana-kanji conversion processing device to which a word extraction method according to an embodiment of the present invention is applied, FIG. 2 is a flowchart showing the operation of the word extraction method according to an embodiment of the present invention, and FIG. It is a figure showing a kanji sound system. 3: Kanji sound separator, 4: Kanji sound device, 5: Input character string attribute determination section, 6: Kanji sound attribute determination section, 7: Searched character string setting section, 8: Dictionary search section, 9: Word dictionary.

Claims (1)

[Claims] (1) In a kana-kanji conversion processing device having means for separating an input character string using kanji sounds and kana as minimum units, a first means for determining the arrangement of kanji sounds and kana in the input character string separated by the means; , the second step determines whether there is a suffix with the same reading as the kanji sound in the character string to be analyzed.
and at least the first or second of the string to be parsed.
If the second minimum unit is a Kanji sound and the third minimum unit is a Kanji sound with the same reading as the suffix, a third means is provided for creating a search string using the three minimum units. A word extraction method characterized by: