JPH03150663A - Sentence elaboration supporting device - Google Patents

Abstract

PURPOSE:To generate the sentence of the Japanese language easy to understand by providing a selection instructing means to select one of plural syntaxes displayed on a display part, and providing a syntax evaluating part to evaluate one syntax selected by said selection instructing means. CONSTITUTION:The selection instruction means to select one of plural syntaxes displayed on the display part 4 is provided at an input part, and the syntax evaluating part 9 to evaluate one syntax selected by said selection instructing means is provided. Then, an operator selects one of the syntaxes displayed beforehand on the display part by the selection instructing means, and sends this selected information to the syntax evaluating part 9, and the complicatedness of the syntax is evaluated by the evaluating part 9 by using the modification relation of the syntax or the number of partial trees in the course of the analysis of the syntax. Thus, the sentence of the Japanese language easier to understand can be generated.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention relates to a draft support device in natural language processing.

Conventional technology Conventional methods for supporting the proofreading of text include the following:
There are various methods as shown in ) to 4).

1) A method of detecting a (suspected) error pattern and issuing a warning in response to it.

b. In addition to issuing a warning,
How to suggest corrections.

c. In addition to issuing warnings.

How to auto-correct.

2) A method of presenting alternative expressions, synonyms, and paraphrases for the specified part of the expression based on the operator's instructions.

3) Based on statistical information such as the number of terms, number of predicates, and ratio of kanji in a sentence, @arning is too long, too complex, etc.
How to issue a warning.

4) A method of supporting checks by operators using automatic index functions, etc.

All of these methods are effective in their own way, but for example, methods 1) and 2) only support local expression, and method 3) does not support how to express it. remains at a statistical level.

Problems to be Solved by the Invention Therefore, in the conventional method described above, when the sentence structure becomes complicated, the degree of complexity is evaluated and the sentence is corrected according to the degree of evaluation. There was no way to support them.

For example, now, as a concrete example, ``I told her I bought it at the department store yesterday. I gave you a ring. ” Suppose there is a sentence.

In this case, since the sentence itself is not an error and has the same meaning, and contains only two verbs, the conventional system outputs it as a normal sentence without receiving any warnings. However, in reality, the above-mentioned sentence would be better expressed as, ``I gave my girlfriend the ring I bought at the department store yesterday.'' The relationship would be clearer, and the sentence would be more Japanese-like and easier to understand. It becomes a sentence. However, conventional deposition support systems do not have such an analysis method, and there is a tendency for problems to occur frequently, especially when sentences become complex.

means to solve the invention Therefore, in order to solve such problems.

The present invention provides a manuscript support device in which an input sentence input from an input unit is parsed by a syntactic analysis unit using a dictionary and phrase structure rules, and the sentence structure as a result of the analysis is displayed on a display unit. A selection instruction means for selecting one of the plurality of sentence structures displayed on the display section is provided, and a sentence structure evaluation section is provided for evaluating the one sentence structure selected by the selection instruction means.

Accordingly, the operator selects one of the sentence structures pre-displayed on the display section using the selection instruction means, sends the selected information to the sentence structure evaluation section, and the evaluation section changes the sentence structure. By evaluating the complexity of a sentence structure using the number of subtrees in the middle of analyzing relationships or sentence structure, it becomes possible to create Japanese sentences that are easier to understand.

Example An embodiment of the present invention will be described based on the drawings.

First, FIG. 1 shows the overall configuration of this device. In this case, the input section 1, syntax analysis section 2, analysis result section 3, display section 4, furthermore, the dictionary section 5 for storage, the phrase structure rule section 6, the message section 7, and the input judgment section 8 have well-known configurations. This embodiment is characterized by the provision of a sentence structure evaluation section 9.

The sentence structure evaluation section 9 evaluates one sentence structure selected by a selection instruction means (not shown). Furthermore, the input determination section 8 determines whether to send the human power information to the sentence structure evaluation section 9 or the syntax analysis section 2 based on the contents of the input information.

Further, FIG. 2 shows the overall flow of processing according to this method, and FIG. 3 shows the flow of evaluation processing in the sentence structure evaluation section 9, which is the main part of the present invention.

In such a configuration, first, the general flow of the apparatus will be described.In the six-manpower section 1, an operator inputs a sentence or a selection instruction means inputs a selection instruction. The input determining section 8 determines the content of the input and determines whether to send the data to the syntactic analysis section 2 or the sentence structure evaluation section 9. Then, when the input sentence is sent to the structure analysis section 2, using the dictionary information and rule information stored in the dictionary section 5 and the phrase structure rule section 6,
Performs an exhaustive search type syntax analysis. A plurality of sentence structures thus obtained are stored in the analysis result section 3, and at the same time, the results are sent to the display section 4 and the contents thereof are displayed.

The operator then observes the displayed content and
Select one of the analysis results you want and input that information into input section 1
input using the selection instruction means.

Thereafter, the input information (code) is sent to the sentence structure evaluation section 9 via the input determination section 8.

This sentence structure evaluation section 9 reads out the selected sentence structure from the analysis result section 3, examines information regarding the structure, selects a message from the message section 7 according to the result, and displays its contents on the display section. .

In this way, by performing a series of tasks, it is possible to create an easy-to-understand sentence that is closer to the Japanese expression than the intended purpose.

Therefore, a specific example of the evaluation method in the sentence structure evaluation section 9 will be described. Assume now that an input sentence as shown in FIG. 4 has been input.

In this case, if, for example, two sentence structures are obtained as a result of the analysis of the input sentence, the two sentence structures as shown in FIG. 5 are displayed on the display section 4.

At this time, assume that what the operator intends is the first sentence structure. In other words, when the operator inputs the first selection instruction with the intention of saying "I bought it yesterday" instead of "I gave it to you as a present yesterday," the input result is sent to the sentence structure evaluation section 9 via the input judgment section 8. It will be done.

Then, in the sentence structure evaluation unit 9, regarding the analysis result corresponding to the first sentence structure, the dependency with the longest distance is a relation, that is, here, the distance 6 (clause ) is the longest, and the result is larger than the predetermined threshold (in this case, 5), so a message indicating that "r dependency" is too far in this case is selected from message s7 and displayed. do. Then, by looking at the contents of the message and performing actions such as modifying the sentence structure, the operator can finally modify the message into an easier-to-understand Japanese sentence as shown in FIG.

Therefore, conventional stacking support methods could not support sentence structures that are grammatical and do not contain errors, as shown in FIG. 5, whereas the method of this embodiment By adopting this method, even if a sentence is grammatically correct, the sentence structure is analyzed to make it easier to understand, and the result is evaluated and a message is sent to the operator to improve the original structure of the sentence. It is possible to realize a concept similar to (not proofreading).

Finally, in this embodiment, the evaluation method in the sentence structure evaluation unit 9 includes, in addition to the evaluation using the distance of "dependency" relationship, for example, It is also possible to evaluate by setting an upper limit on the number of In this case, the number of subtrees created when each word is input, for example,
In the first sentence structure in Figure 5, when the word ``ring'' is manually written, three subtrees have been created: ``For my girlfriend'' and ``I am J. ``It's the ring I bought at the department store yesterday.'' , the number of subtrees is 3, and such evaluations are examined in order from the beginning of the sentence, and the upper limit value is compared with a threshold value to select a message.

Effects of the Invention The present invention provides a drafting support device in which an input sentence input from an input unit is parsed by a syntactic analysis unit using a dictionary and phrase structure rules, and the sentence structure as a result of the analysis is displayed on a display unit. , a selection instruction means is provided for selecting one of the plurality of sentence structures displayed on the display section, and a sentence structure evaluation section is provided for evaluating the one sentence structure selected by the selection instruction means. Therefore, the operator uses the selection instruction means to select - from among the sentence structures displayed in advance on the display section, and sends the selection information to the sentence structure evaluation section. By evaluating the complexity of a sentence structure using the number of subtrees during structure analysis, it is possible to support the creation of Japanese sentences that are easier to understand than before.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing an embodiment of the present invention, Fig. 2 is a flowchart showing the overall flow, Fig. 3 is a flowchart showing the flow of evaluation processing, and Fig. 4 shows an example of an input sentence. FIG. 5 is an explanatory diagram showing an example of the result of syntax analysis displayed on the display unit, and FIG. 6 is an explanatory diagram showing a sentence obtained as a result of correcting the input sentence. 1... Input sentence, 2... Syntactic analysis section, 4... Display section, 9... Sentence structure evaluation section J, I Figure

Claims (1)

[Claims] In a drafting support device, an input sentence input from an input unit is parsed by a syntactic analysis unit using a dictionary and phrase structure rules, and the sentence structure as a result of the analysis is displayed on a display unit,
A selection instruction means for selecting one of the plurality of sentence structures displayed on the display section is provided, and a sentence structure evaluation section is provided for evaluating the one sentence structure selected by the selection instruction means. Featured manuscript support device.