JPS584481A - Japanese word analyzing system - Google Patents

Abstract

PURPOSE:To realize the processing of significance of a Japanese word by means of a computer, for example, mechanical translation and questionnaire, by expressing the significance expressed in Japanese through the connection of phrases and clauses mechanically. CONSTITUTION:A phrase/clause synthesizer 1 splits a sentence into phrases/ clauses and a connection analyzer 2 receives them and outputs a list among the phrases and clauses. In the device 1, a series of words is inputted to a synthesizer 7 for clauses/phrases, whether each word is an independent word or an adjective word at a discriminator 8 to synthesize th words into a sentence. In the device 2, when two clauses are inputted, an extracting device 9 determines their grammatical characters. A discriminator 4 checks the presence/absence of grammatical modification between the clauses, and if present, the clauses are transferred to an extracting device 10. If absent, the are given to a storage device 11. The device 10 seeks the number of a functional word and conceptional word constituting the clauses, and gives the number to a discriminator 5. The device 10 outputs a list between the connection of clauses when the presence of connection is returned.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a Japanese language analysis system that is capable of inputting a Japanese sentence, analyzing its meaning, and outputting information indicating relationships between clauses.

In conventional Japanese language analysis systems, each verb requires an action agent, an action partner, and an object. However, in Japanese, expressions related to action, the opponent of the action, and time are often omitted, so if you try to prepare an accurate case pattern, you will end up with (a) a complicated form. (cyJ Many case patterns are required for one verb.

The conventional method has this drawback.

The present invention aims to eliminate the above-mentioned drawbacks, and is to examine the relationships between Japanese phrases using a dictionary with a relatively simple structure, and to output information on the relationships between Japanese phrases. The purpose is to provide a word analysis system.

Therefore, the Japanese language analysis system of the present invention includes a grammatical relationship discriminator that determines the presence or absence of grammatical relationships between multiple types of clauses, a grammatical relationship discriminator that determines the presence or absence of semantic relationships between multiple types of independent words and multiple types of attached words, and A semantic relationship discriminator that determines the presence or absence of a semantic relationship between multiple types of pairs of independent words and adjunct words and independent words; converts an input Japanese sentence into a string of words using the Japanese dictionary search device described above; A phrase synthesis device that combines the phrases into a string of phrases after decomposition, and a grammatical relationship determining device that determines whether or not there is a grammatical relationship between two phrases included in the phrase string created by the phrase synthesis device. If it is determined that there is a grammatical relationship, the presence or absence of a semantic relationship between the two clauses is checked using the above semantic relationship determination device. The modification that outputs the clauses and the clause quantity relationship is characterized by being equipped with an analysis device. Hereinafter, the present invention will be explained with reference to the drawings.

Fig. 1 is a diagram showing an overview of an embodiment of the present invention, Fig. 2 is a block diagram of an embodiment of a phrase synthesis device, Fig. 3 is a diagram illustrating a Japanese dictionary search device, and Fig. 4 is a self-supporting device. FIG. 5 is a block diagram of one embodiment of the analysis device; FIGS. 6 and 7 are diagrams explaining the semantic relationship discriminating device; FIG. 1 is a diagram explaining the grammatical relationship discriminating device, and FIG. 1O is a diagram explaining the flow of processing.

The present invention is based on the recognition that 1. The meaning expressed in Japanese can be understood by the relationship between bunsetsu and bunsetsu, and it extracts the characteristics of bunsetsu grammatically and semantically. , the relationships between clauses are examined by separating them into grammatical and semantic aspects.

FIG. 1 shows an overview of one embodiment of the present invention,
1 is a phrase synthesis device, 2 is a dependency analysis device, 3 is a Japanese dictionary search device, 4 is a grammatical relationship discrimination device, and 5 is a semantic relationship discrimination device.

Japanese sentences do not have the custom of separating sentences, so phrase synthesis device 1 is used.
refers to the Japanese dictionary search device 3 or the like and divides the input Japanese sentence into a string of clauses. The dependency analysis device 2 examines the relationships between clauses with reference to the Japanese dictionary search device 3, the grammatical relationship determination device 4, and the semantic relationship determination device 5, and outputs an inter-clause relationship table.

FIG. 2 is a block diagram of one embodiment of the clause synthesis device 1, in which 6 indicates a word unit division device, 7 indicates a clause synthesis device, and 8 indicates an independent word/adjunct word discrimination device. .

Note that the same reference numerals as in FIG. 1 indicate the same parts. The Japanese dictionary search device 3 has a Japanese dictionary as shown in FIG. 3, and this dictionary has a plurality of entries in which words and word relationship information are written. The independent word/adjunct word discriminating device 8 has an independent word/adjunct word discriminating dictionary as shown in FIG. There is. Assume that the Japanese sentence ``I gave her a book.'' is input. The word-by-word division device 6 uses the Japanese dictionary search device 3 to search for words corresponding to the input sentence while dividing the input sentence into words. After processing up to this point, the input sentence becomes [I U has girlfriend lol K
LJ Honguchi gave Utaguchi. ”.

The word string divided into unit words is input to a phrase synthesizing device 7. In this case, each word is examined to see whether it is an independent word or an adjunct word using an independent word/adjunct word discriminator and synthesized into a clause. -Given.'' is combined into a string of clauses.

FIG. 5 is a block diagram of one embodiment of the analysis device 2. As shown in FIG. In FIG. 5, 9 is a grammatical feature extraction device, 10
11 indicates a semantic feature extraction device, and 11 indicates an unresolved phrase storage device. Note that the same reference numerals as in FIG. 1 indicate the same parts. The grammatical relationship determining device 4 has a table dictionary as shown in FIG. The example in FIG. 8 shows that there is a grammatical relationship (a dependency relationship) between the predicate modification type clause and the verb clause. The semantic relationship determination device 5 is.

It has a matrix-type semantic relationship dictionary as shown in FIG. The horizontal part of this matrix is divided into function words and concept words, and the vertical part is divided into concept words and concept word/machine language pairs. A function word can be thought of as an attached word, which is attached to an independent word to form a clause and determines the attributes of the independent word in that clause.A function word is a part of an attached word. While representing the concept words, the relationship between the concept words can also be classified in detail. 7th
The figure shows how the rows and columns of the semantic relationship dictionary matrix are divided and used, and what the parts in the matrix represent. In Figure 6, if we look at the intersections with function words for "bird" in the vertically arranged conceptual words, we find "living thing/actor" and "owned car".
is "1", which indicates that the attributes of "bird" include "creature/actor" and "owned car." On the other hand, the intersection with "1 hour" is "0", which indicates that "bird" does not have the attribute K "time". The word ``(ga)'' is written after ``creature/actor,'' which indicates that the particle ``ga'' represents this attribute.

Next, the vertically arranged concept words and function words ``bird x living thing/actor''
If we look at the intersection of conceptual words for ``eat'', ``
1". This shows that there is a relationship with ``eating'' through the main attributes of ``birds'' as creatures and movements.

``0'' for ``one other bird'' and ``bird'' is
It shows that there is no relationship through the attribute of creature/actor, but the zero-order ``bird x owner'' shows that there is a relationship with ``feathers'' through the concept of ownership.

The rdJ between "bird x owner" and "bird" indicates that the relationship has not yet been defined. Furthermore, as shown in Figure 7, the reason that the "conceptual word" and "conceptual word" parts are not used is that there is no direct relationship between the two conceptual words. This is because it is based on the idea that "there exists." The reason why the "concept word x function word" and "function word" parts are not used is that there is no meaning in setting them.

When the phrase string ``Bird eats U'' is input, the following processing is performed to check whether there is a semantic relationship or not. i, “Tori” No. K1000,
The number 10 and number 11 are seen for ``ga'', and the number 2000 is used for ``eat'', and the correspondence between words and numbers can be found by looking in a Japanese dictionary. .

One horizontal row corresponding to "bird" (1000th horizontal row) and "
Vertical columns corresponding to "ga" (10th and 11th vertical columns)
When I checked the intersections of
If the intersection with the number is "0", proceed to the first-order processing.0 If both are "0", there is no semantic relationship. Check the intersection between the horizontal row and the vertical row (2000th) corresponding to "eat". If the intersection point is "1", there is a #i1 relationship, and if it is "0", there is no #i1 relationship.

-Next, the person in charge will explain the operation of the analysis device 2. In the analysis device 2, the engagement begins to operate from the time when two phrases of one phrase are sent.

First, the grammatical feature extraction device 9Vc determines the grammatical features of each clause using the Japanese dictionary search device f13.

For example, in the phrase ``wa'', ``wa'' is a pronoun and ``wa'' is a topic-presenting particle. Therefore, "I am" becomes a one-pronoun topic-presenting particle, that is, a predicate-modifying clause. Once the grammatical features of the two clauses □ are determined, the grammatical relationship discriminator 4 is used.

It is checked whether there is a grammatical modification relationship between these two clauses. If one grammatical modification relation exists, control is transferred to the semantic feature extraction device 10, and if one grammatical modification relation does not exist, the clauses are sent to the unresolved clause storage unit fllF11. Unresolved clause storage device 1
1, the phrases that are sent are stored in that order. and.

When a phrase is requested, the most recently sent phrase is handed over first. Semantic feature extraction device 10
refers to the Japanese dictionary search device 3 to find the numbers of the concept words (independent words) and function words (attached words) that make up one clause. Next, the two clauses represented by the word numbers are is passed to the physical relationship discriminator 5.

Request an investigation to determine whether there is a semantic relationship. When this request is received, the semantic relationship determination device 3 determines whether there is a semantic relationship as described in connection with FIGS. 6 and 7, and returns the presence or absence of a semantic relationship.

If there is a semantic relationship, there is not only a semantic relationship, but also
The word number of the function word is also returned. When a semantic relationship is returned, the semantic feature extraction device 10 outputs a clause amount relationship table assuming that the relationship between the two clauses has been determined. and.

requests a new phrase from the unresolved phrase storage device 11;
Let this be the new relevant phrase. The uke uses the same clauses as before, and the presence or absence of grammatical and semantic relationships is checked as before.This is repeated until the unresolved storage device 11 is empty. On the other hand, if there is no semantic relationship, the two clauses are unresolved in memory! 1tllK sent. In the above process, after sending the phrase to the unresolved phrase storage device 11, assuming that there is no grammatical or semantic relationship between the phrases, a new phrase is requested from the phrase synthesizer i<.
The grammatical and semantic relationships are examined again between the phrases sent from the phrase synthesizer 1 and the phrases retrieved from the unresolved phrase storage device 11. Figure 9 shows the input of the Japanese language analysis system. This shows an example of output. "I gave her a book."
When inputting the Japanese sentence into the Japanese language analysis system S, 1 is obtained.
An inter-clause relationship table T as shown is output.

Figure 1O) through d) show the flow of processing in the Japanese language analysis system of the present invention. As shown in Figure 10 (a), the 9th phrase “I am” and the phrase “to her”
When the phrase ``Watashi wa'' is input to the grammatical feature extraction device 9, the grammatical feature extraction device 9 uses the Japanese dictionary search device 3 to determine that ``Watashi wa'' is a predicate-modifying phrase.

Learn that ``to her'' is a verb-modifying clause.

The grammatical feature extraction device 9 requests the grammatical relationship determination device 4 to investigate whether there is a grammatical relationship between the predicate modification type clause and the verb modification type clause. The grammatical feature extraction device 9 sends these phrases to the unresolved phrase storage device 11 and requests new phrases.

As shown in FIG. 10 (B), the phrase "Kojo k" is taken out from the unresolved phrase storage device 11, and the phrase "Book" is sent from the sentence synthesis device 1.0 Grammatical feature extraction device 9 is "to her" using 1 Japanese dictionary search device 3
Examine the characteristics of the phrase ``'' and the phrase ``book''. When it is determined that ``to her'' is a verb-modifying clause and ``Ichihon wo'' is a verb-modifying clause, the grammatical feature extractor 9 uses the grammatical discriminator 4 to determine whether or not there is a grammatical relationship between the two. investigate.

If it is determined that there is no grammatical relationship, these clauses are sent to the unresolved clause storage device 11 and a new front clause is requested.

As shown in FIG.

The grammatical feature extraction device 9 checks whether there is a grammatical relationship between the two as described above. If a grammatical relationship is found, data and control are passed to the semantic feature extractor 10. The semantic feature extraction device 10 uses a Japanese dictionary search device 3. “O” in Tasoshi@5”
indicates the object.

The "wo" in @6# indicates the place to pass. The semantic feature extraction device 10 uses the semantic relationship determination device 5 to
The presence or absence of a semantic relationship between the 10th conceptual word and the 5th function word and the presence or absence of a semantic relationship between the 10010th conceptual word and the 6th function word are examined.

When it is determined that there is a semantic relationship between the concept word No. 10010 and the function word No. Check whether there is a semantic relationship between 10010 and 10010, as mentioned earlier.
You can find out by checking whether the intersection of the horizontal row of +5Xn and the vertical row of number 30000 is "1" or "0". If it is determined that there is a semantic relationship.

The relationship between clauses 1 and 2 in the 0-clause amount relation table, which outputs the inter-clause relation table as shown in the figure, is "object", but this is because the function word number 15# is changed to a form that is easy for humans to understand. It was converted into .

As shown in Figure 10 (b), the clause ``to her'' and the phrase ``
When the phrase "I gave" is input, the grammatical feature extractor 9 checks whether there is a grammatical relationship between the two phrases as before.0 If there is a grammatical relationship, the grammatical feature extractor 9 checks the meaning. Control and data are passed to the target feature extraction device 10. The semantic feature extraction device 10 is.

Using the Japanese dictionary search unit 3 and the semantic relationship determination device 5, we examine whether or not there is a semantic relationship between the phrase ``to her'' and the phrase ``4 eta''. Since there is a semantic relationship, an inter-clause relationship table as shown in Figure 1 is output.

As is clear from the above explanation, 1.According to the present invention, 1.The meaning content expressed in Japanese is mechanically expressed by the relationship between clauses. It makes it possible to realize a response system.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing an outline of one embodiment of the present invention, FIG. 2 is a block diagram of one embodiment of a phrase synthesis device, and FIGS.
FIG. 8 is a diagram for explaining a semantic relationship discriminating device, and FIG. 8 is a diagram for explaining a grammatical relationship discriminating device. FIG. 9 is a diagram showing an example of input and output of the Japanese language analysis system. FIGS. 10A, 10B, 10C, and 10C are diagrams illustrating the flow of processing. 1...Phrase synthesis device, 2...Kaishi-uke is an analysis unit. 3...Japanese dictionary search device, 4...grammatical relationship discrimination device, 5...semantic relationship discrimination device, 6...device for dividing into word units, 7...device for synthesizing into phrases , 8... Independent word/adjunct word discrimination device, 9... Grammatical feature extraction device,
10... Semantic feature extraction device, 11... Unresolved phrase storage device. Patent Applicant: Fujitsu Limited Representative Patent Attorney Yotsube Kyotani

Claims (1)

[Claims] Japanese dictionary search device, grammatical relationship determination device that determines the presence or absence of grammatical relationships between multiple types of clauses, semantic relationships between multiple types of independent words and multiple types of attached words, and independent words and attached words A semantic relationship determination device that determines the presence or absence of a semantic relationship between multiple types of and independent words, and a device that analyzes an input Japanese sentence into a string of words using the Japanese dictionary search device described above, and then divides it into a string of phrases. and a phrase synthesis device that synthesizes a sentence into a phrase synthesizer, and the grammatical relationship determination device that checks whether there is a grammatical relationship between two phrases included in the string of phrases created by the phrase synthesis device. In the case of a tosaraina, the presence or absence of a semantic relationship between the two phrases is checked using the above semantic relationship determination device, and if it is determined that there is a semantic relationship, the relationship between the two phrases and the amount of phrases is determined. A Japanese language analysis system characterized in that a dependency that outputs an analysis device is provided.