JPS60200359A - Simple sentence producer - Google Patents

Abstract

PURPOSE:To facilitate easy natural language processing and its debug for collection, analysis, etc. of data by providing a word dividing means, a modification analyzing means, etc. and decomposing a natural language sentence containing complicated factors tangled with each other into a simple style. CONSTITUTION:A sentence supplied from an input part 2 is divided into words through a word dividing part 3 and then collated with the words stored in a dictionary 6 stored in a memory. Thus a word list is produced and sent to a paragraph composing part 4, and the part 4 supplies the word list every word at and after the head of the sentence. A paragraph list thus produced is sent to a modification analyzing part 5, and an analysis tree is produced via the dictionary 6 and an indicating part 7. This analysis tree is sent to single sentence extracting part 8 for production of a single sentence. Thus a complicated sentence is changed into a simple style via those parts 3, 4, 5, 7 and 8 as well as the dictionary 6. This can facilitate easy natural language processing and its debug for collection, analysis, etc. of data.

Description

DETAILED DESCRIPTION OF THE INVENTION (1) Technical Field of the Invention The present invention relates to natural language processing by a data processing device, and relates to a simple sentence generation device that generates a simple sentence from an input sentence.

(2) Prior art and problems Natural language sentences generally have a very complex structure, so it is extremely difficult to use them directly as data or collect data from the sentences when performing computer processing. Have difficulty. However, even if the sentence has a complex structure, it can be called a simple sentence (a sentence consisting of one subject and one predicate).
By decomposing it into , it becomes easier to determine the syntactic pattern and the case of the verb because there is less syntactic ambiguity, which makes it easier to process by computer. However, in order to break it down into simple sentences, the structure of the input sentence must be determined, and in order to determine that structure, the syntactic pattern or the case of the verb must be known. However, since we are trying to find a simple sentence to determine the origin, this syntactic pattern, or the case of the verb, there is a problem in that it is impossible to do this automatically.

Therefore, conventionally, when processing natural language using a computer (data collection), a method has been adopted in which the target sentence is manually input into a simple format, or a simple format is selected and input. However, the former method required a great deal of effort, and the latter method had the disadvantage that important information could be leaked.

(3) Purpose of the Invention In view of the above-mentioned drawbacks of the conventional art, the present invention decomposes the sentence meaning of natural language involving complex factors into a simple format, thereby facilitating natural language processing such as data collection and analysis, and its debugging. The purpose is to provide a method that can be used.

(4) Structure of the Invention The purpose of this invention is to provide a phrase synthesis means that divides input text data into words to form a word list, and connects the word restructuring in phrase units to generate a phrase list '6. a dependency analysis means for analyzing the dependency relationship between the clauses in the text; a simple sentence extraction means for creating an analytic tree in response to the output of the means and extracting a simple sentence from the analytic tree;
The dependency analysis means determines whether or not a clause located in the middle of three consecutive clauses in the clause list relates to the following clause, with reference to the grammar table on the storage device, and When the relationship between three clauses is specific, this can be achieved by determining the dependency relationship externally.

(5) Embodiment of the Invention FIG. 1 is a block diagram of an example of a device implementing the present invention, in which 1 is a simple sentence generation device, 2 is an input section, 3 is a word segmentation section, 4 is a phrase synthesis section, Reference numeral 5 represents a dependency analysis section, 6 a storage device, 7 an instruction section, and 8 a simple sentence extraction section.

In FIG. 1, a sentence input via the input unit 2 is compared with words stored in a dictionary registered in the storage device 6 in the word division unit 3 to create a word list, and then sent to the phrase synthesis unit 4. passed on.

The phrase synthesis unit 4 inputs the word list one word at a time from the beginning of the sentence,
Phrase synthesis is performed by the following operations. The processing flow is shown in FIG.

(b)・・・・・・・・・Things with consecutive nouns are
Connect this.

(b)・・・・・・Particles connect to the previous word.

(c)・・・・・・Auxiliary verbs connect to the previous word.

ni) ......Formal nouns that have no meaning (for example, 6 oki'', ``koto'', etc.) are connected to the previous word.

)・・・・・・・・・Adnominals are connected to the word on the right.

(to)・・・・・・・・・The verb ending connects to the previous word.

(g)・・・・・・Commas, periods, and dots connect to the previous word.

)・・・・・・Connect everything inside the parentheses. Also, attach the whole thing enclosed in parentheses to the previous word.

ω)・・・・・・・・・Unit (bit, byte, α, H
) connects to the previous word.

Leave the grammatical attributes of the last word in the clause when concatenating.

For example, 1 in 1 school (continuous modification) 1 In the example of (noun modification) 1 study 1, Grammatical attribute of “no (noun modification)” money 1 1 School (noun modification) 1.

The attributes of a clause include a noun, a verb, a noun modification, a verb modification, an adjunctive modification verb, and an adnominal modification verb. Among these, noun modifications and verb modifications may co-occur, such as "and", and the verb at the end of a sentence has the attribute "1 sentence end".The attributes of each of these clauses include particles and Depending on the ending of the word, for example, particles such as ``, -zo, 1'' modify the verb, ``no'', ``in'', etc. modify the noun, and conjunctive verbs such as ``6'' modify the verb. .

When the phrase list created by the phrase synthesis section 4 using the above method is sent to the dependency analysis section 5, the dependency analysis section 5 uses the "three analysis windows" to analyze the meaning from the end of the sentence. Let's do it. ``Three analysis windows'' refers to three consecutive frameworks and overlapping sentences so that one clause fits in each framework, and is used to facilitate explanation of analysis. From the left, open each analysis window (framework) with a
Name it MXR. Then, an analysis tree is created by performing all the analysis lines by combining the attributes of the clauses in each window. The analysis procedure using the "three analysis windows" is as shown in Table 1 and Figure 3. The combination of analysis procedures in Table 1 is storage device 6.
It is stored as a grammar table inside. If a user's instructions are required during analysis, message output or instructions from the user are performed via the instruction section 7.

The parse tree created in this way is sent to the simple sentence extraction section 8, where a simple sentence is generated. The simple sentence division at this time is performed in the following steps.

Table 1■・・・・・・Extract a noun phrase (including adnominal modifications) and leave only the last word.

■・・・・・・・・・Divide the conjunctive modification.

■・・・・・・Converts the continuous form to the final form and outputs a simple sentence.

The process of generating a simple sentence in each part above will be further supplemented with specific examples as follows.

For example, when the input sentence is ``Graphic information is given for X-Y coordinates, and this coordinate area is called a screen,'' the word list is ``Graphic 1 information 1 is given by 1 for IX-Yl coordinates 1. 1.1 this 1 coordinate 1 area 1 gold 1 screen 1 and 1 call 1 1゜'', and the phrase list is [Graphic information is IX-Y
1 is given for the coordinates, 1 this coordinate area is 1 screen and 1
call. 1" (1 indicates a break between words or phrases).

Figure 4 is a diagram showing the process of further analyzing the clause list using the "three analysis windows", where 9 is the input sentence, 10 is the clause break, and 11 to 13 are the "three analysis windows". is Ll
12 represents M113 represents R. As a result of such analysis, an analytic tree is obtained.

Figure 5 shows the parse tree and extracted simple sentences.
e) represents a parse tree, and (d) represents a simple sentence.

(6) Effects of the Invention As explained in detail above, the simple sentence generation device of the present invention can be realized with a simple configuration, and it is possible to refer to the user for parts that are difficult to judge mechanically. Since the system follows the instructions given by the computer, it has the advantage of requiring less dictionary information and the like, and the loss of processing time is small, which is very effective.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of an example of a device implementing the present invention, Fig. 2 is a flowchart showing the operation of the phrase synthesis section, Fig. 3 is a flowchart showing the operation of the dependency analysis section, and Fig. 4 is a clause list. of"
Diagram 5 showing the process of analysis using “Three Analysis Windows”
The figure shows the parse tree and the extracted simple money. 1... Simple sentence generation device, 2... Input section,
3... Word division unit, 4... Clause synthesis unit, 5... Dependency analysis unit, 6... Dictionary, 7... instruction section, 8... simple sentence extraction section,
9... Input sentence, 10... Clause break, 11-13... "Three analysis windows" Figure 1 Figure 2 Figure 3

Claims (1)

[Claims] A phrase synthesizing means that divides the input text data into words to form a word list, and connects the word list in phrase units to generate phrase restructuring; and a section that analyzes the dependency relationship between phrases in the phrase list. a receiver parsing means, an instruction receiving means for externally determining the dependency relationship between clauses, and an inserting means; The parse tree has simple sentence extraction means for extracting simple sentences, and the dependency analysis means determines whether a phrase located in the middle of three consecutive phrases in the phrase list is related to a subsequent phrase. 1. A simple sentence generation device characterized by making a judgment by referring to a grammar table on a storage device and, when the relationship between the three clauses is a specific one, to externally judge the relationship between the three clauses.