JPH0785057A - Syntax analyzing method and device therefor - Google Patents

Abstract

PURPOSE:To improve a syntax analysis rate qualitatively by performing KANA (Japanese syllabary)/KANJI (Chinese character) conversion of a character string in an inputted sentence, replacing the character string in the sentence with a converted character string mixing KANA and KANJI, and providing a process fro performing syntax analysis again. CONSTITUTION:While referring to a KANA/KANJI conversion dictionary 23, a sentence is converted to a KANJI mixing sentence, and the word of 'kaisha' (in KANA meaning 'company') is converted to 'kaisha' (in KANJI), for example. The syntax analysis is performed by replacing this converted character string 'kaisha' (in KANJI) with the character string of 'kaisha' (in KANA) in a sentence as the original analysis target. First of all, a sentence analysis program 24 possesses one part of a sentence to be analyzed from an external storage device. When the entire attention sentence is completely analyzed, the syntax analysis is performed next. Even when any part mixing KANA/KANJI conversion exists in the sentence to perform the syntax analysis as the result of this processing, the influence to be exerted upon the syntax analysis by that part can be reduced, and the analysis rate of syntax analysis can be improved.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a syntax analysis method and apparatus, and more particularly to a syntax analysis method and apparatus for analyzing the syntax of a document in an existing Japanese document file.

[0002]

2. Description of the Related Art Today, various attempts have been made to analyze a sentence composed of natural language. The syntactic analysis itself is a means used for, for example, interlingual translation (for example, translation from Japanese to English), voice recognition, voice output, and the like, and although the itself is not the final purpose, the above processing is performed. It is essential to do.

[0003]

Now, let's assume that a sentence created by a word processor device or a program therefor, which is widely used today, is parsed. In the syntax analysis, the minimum unit forming a sentence is cut out, and the relationship between the cut out minimum units is analyzed based on the morpheme dictionary. Here, since the sentence is stored in a storage medium such as a floppy disk, the characters are sequentially read out from the sentence and the syntax is analyzed. The problem here is that if the input character or character string is not in the dictionary for morphological analysis, the analysis process for the focused character or character string will fail. As for characters or character strings that are not in the morphological analysis dictionary, "Hiragana" is often the place that should be originally written in Kanji.

[0004]

SUMMARY OF THE INVENTION The present invention has been made in view of such a conventional technique, and even if a character string originally represented by Kanji remains in Hiragana for some reason, the qualitative analysis rate is high. It is an object of the present invention to provide a syntax analysis method and a device capable of performing various improvements. In order to solve this problem, the syntax analysis method of the present invention includes the following steps. That is, a parsing method for parsing a Japanese sentence, wherein if the parsing of a given sentence fails, a process of converting kana-kanji to the hiragana character string in the input sentence, and a kana-kanji character by the process. When converted into a mixed character string, the step of replacing the corresponding hiragana character string portion with the converted mixed kana-kanji character string and re-performing syntactic analysis.

The syntax analysis device of the present invention has the following configuration. A parsing device for parsing a Japanese sentence, wherein when the parsing of a given sentence fails, means for converting kana-kanji to the hiragana character string in the input sentence and kana-kanji mixed characters by the means. When converted into a string, the corresponding hiragana character string portion is replaced with the converted kana-kanji mixed character string, and the syntax analysis is performed again.

[0006]

In the process or configuration of the present invention, if the syntactic analysis fails, kana-kanji conversion is performed on the hiragana character string in the sentence. Then, the kana-kanji mixed character string obtained by the kana-kanji conversion is used to replace the corresponding hiragana character string portion, and the syntax analysis is performed again.

[0007]

Embodiments of the present invention will now be described in detail with reference to the accompanying drawings. FIG. 1 is a block diagram of the natural language processing system of the embodiment. In the figure, 1 is a CPU that controls the entire system, and 2 is a ROM that stores a boot program of the CPU 1 and character fonts.
Reference numeral 3 is a RAM used to load various programs and used as a work area of the CPU 1. 4 is a keyboard for giving characters or various processing to this system, 5 is a VRAM for developing a character pattern to be displayed,
A display device 6 displays a video corresponding to the image expanded in the VRAM 5. Reference numeral 7 denotes an external storage device including a floppy disk drive, a hard disk drive, etc., in which various information as shown in FIG. 2 is stored and held as a file.

In FIG. 2, reference numeral 21 is a system program (OS) loaded in the RAM 3 when the power of the system is turned on, 22 is a morpheme dictionary for parsing, and 23 is a normal document edit. A kana-kanji conversion dictionary used in, etc., 24 is a document analysis program for performing the process of the flowchart of FIG. 4 described later, and 25 is a document file created by the user.

The operation contents in the above configuration will be described below. FIG. 3 shows an example of a sentence to be analyzed. The syntactic analysis itself is performed by the morpheme dictionary 22, but since it does not have the characteristics of the present invention itself, the syntactic analysis itself also analyzes the relationship between individual words and is generally known. Since this is the case, the description thereof is omitted. Now, paying attention to the sentence shown in the figure, it will be apparent that the parsing is normally performed for the first character string "I" or the particle "I". The problem is,
The following string "Kaishani" is Hiragana, so it is expected that parsing will fail.

In this embodiment, the kana-kanji conversion dictionary 23 is used to perform kana-kanji conversion.
In view of the current Kana-Kanji conversion, it is a well-known fact that the conversion efficiency of batch conversion is still at a considerable level even though the conversion efficiency of batch conversion is still high. Therefore, also in this embodiment,
This character string "Kaisha ni" is passed as an argument to the Kana-Kanji conversion program (resident in the RAM 3 when the system is started), as in the case of reading by using a normal keyboard. The kana-kanji conversion program also receives a character string as this reading, converts it into a kanji-mixed sentence by referring to the kana-kanji conversion dictionary 23, and returns the conversion result to the request source. As a result,
"Kaisha ni" will be converted into "company" as intended.

In the embodiment, the converted character string "company ni" is replaced with the character string "kaisha ni" in the sentence which was originally the object of analysis, and the syntactic analysis is performed. If the system is a Japanese-English translation device, for example, the syntactic analysis result will be passed to the next Japanese-English translation process as information in which the relationship between individual words is applied to the English grammar. However, since the Japanese-English translation process itself does not have the characteristics of the present invention, the description thereof will be omitted.

If Kana-Kanji conversion also fails, the parsing of the sentence of interest, in principle, fails.
It can be easily inferred from the above description that the failure rate can be at least reduced as compared with the prior art. In addition, when the kana-kanji conversion is performed by reading the character string "Kaisha" or "Kaisha ni", it will be converted to "company" or "company" with almost 100% probability, but there are multiple homonyms. In such a case, the first candidate of the conversion result by the Kana-Kanji conversion dictionary 22 used by the user is set as the conversion result. However, the user may appropriately select one from the conversion candidates. In this case, at least the entire sentence that is being parsed is displayed (including the sentences before and after it if possible), and which hiragana character is the candidate is clearly displayed (if highlighted or colored. Good) This is because when there are a plurality of homonyms, it cannot be judged only by reading them.

In order to specifically execute the above processing, the document analysis program 24 of the embodiment proceeds according to the procedure shown in FIG. First, in step S1, one sentence in the sentence to be analyzed is acquired from the external storage device 7 (or from the RAM 3 when the sentence is in the RAM 3). In step S2, an attempt was made to acquire a sentence, but whether or not it was possible, in other words, the end of the analysis target (EOF: End Of
ile) is detected. If it is determined that the analysis processing for the entire sentence of interest has ended, this processing ends, and if it is determined that one sentence has been earned, the processing proceeds to step S3.

In step S3, syntax analysis is performed. This result determination is performed in the next step S4, and when it is determined that the analysis is successful, the process proceeds to step S5,
Save the analysis result of the sentence of interest as a file,
If any processing is to be performed based on the analysis result, the processing is passed to that processing, and the processing returns to step S1 in order to analyze the next sentence. On the other hand, when it is determined in step S4 that the parsing has failed, the process proceeds to step S6, and it is determined whether or not the character string that is the cause of the parsing failure in the sentence of interest includes a hiragana character string. To do.

If the hiragana character string exists, the process proceeds to step S7, and kana-kanji conversion is performed based on the character string. As a result, if the kana-kanji conversion is successful, the character string of this conversion result is replaced with the corresponding hiragana character string in the sentence of interest, and the syntax analysis of step S3 is performed again. When there are many hiragana character strings and the parsing fails, the above processing is looped for the number of hiragana character strings.
Therefore, the Kanji occupancy rate of one sentence, which is the focus of attention during this period, will gradually increase. In step S6, if the reason why the syntactic analysis fails is not due to the hiragana character string, or if the kana-kanji conversion fails even if it is the hiragana character string, the syntactic analysis for the sentence of interest is performed. Perform failure processing. For example, a process of outputting the analyzed content is performed even if it is unknown, and the process returns to step S1.

As a result of the above processing, even if there is a portion in the sentence to be parsed that has been forgotten to be converted to Kana-Kanji, the effect on the parse analysis due to it can be reduced, and the parse analysis rate is improved. It is possible to let Although the input destination of the sentence to be analyzed is described as the external storage device in the embodiment, any input destination may be used as long as the sentence is sent as a character code string (for example, online. However, the present invention is not limited to these points.

[0017]

As described above, according to the present invention, even if the character string originally represented by Chinese characters remains in Hiragana for some reason, the parse rate can be improved qualitatively. become.

[Brief description of drawings]

FIG. 1 is a system configuration diagram in an embodiment.

FIG. 2 is a diagram showing an internal structure of an external storage device of the embodiment.

FIG. 3 is a diagram for explaining an outline of processing related to syntax analysis.

FIG. 4 is an overall flowchart of a syntax analysis process according to an embodiment.

[Explanation of symbols]

 1 CPU 2 ROM 3 RAM 4 keyboard 5 VRAM 6 display device 7 external storage device

Claims (2)

[Claims] 1. A parsing method for parsing a Japanese sentence, wherein when the parsing of a given sentence fails, kana-kanji conversion is performed on the hiragana character string in the input sentence, A syntactic analysis method, comprising the step of replacing the corresponding Hiragana character string portion with the converted kana-kanji mixed character string when converted into a kana-kanji mixed character string by the step, and performing syntactic analysis again. 2. A syntactic analysis device for syntactic analysis of a Japanese sentence, comprising means for performing kana-kanji conversion for a hiragana character string in an input sentence when the syntactic analysis for a given sentence fails. A syntactic analysis device comprising means for replacing a corresponding hiragana character string portion with the converted kana-kanji mixed character string when the character string is converted into a kana-kanji mixed character string by means, and performing syntax analysis again.