JPS60157659A - Japanese language analyzing system - Google Patents

Abstract

PURPOSE:To analyze exactly a sentence of Japanese language by executing a determination of a related destination of a sentence clause by deciding a coincidence of the sentence clause, a particle of the case of a case pattern, and a meaning. CONSTITUTION:An input sentence 200 is inputted to a word dividing part 1, and the dividing part 1 utilizes word information accumulated in advance in a word dictionary storage part 4, divides the input sentence 200 into trains of a word unit and sends it to a sentence clause assembling part 2. The assembling part 2 uses grammar information in a sentence clause, stored in an in-sentence clause grammar information storage part 5, basing on a received word train data 210, and assembles the word train data 210 to plural sentence clause units. A relation receiving and analyzing part 3 uses case pattern information of a term stored in a case pattern dictionary storage part 6, analyzes a sentence train data 220 sent from the assembling part 2, decides a related reception of a sentence clause and outputs an analysis result 230.

Description

[Detailed Description of the Invention] [Field of Application of the Invention] The present invention relates to a method for analyzing Japanese text, and particularly relates to a Japanese language analysis method that performs all the analysis of clauses and their dependencies by using them within case patterns of predicates. .

[Background of the invention]

Focusing on the conventional case pattern, that is, the unique case-dominant structure of Japanese pragmatics, we use the grammatical and semantic constraints of Japanese pragmatics and the words that are dominated by them, especially nouns. There is a method to analyze Japanese sentences. This is a method for analyzing the relationship between a single predicate and the noun clause related to it, and is basically an analysis method for simple sentences.

However, the sentences that appear in actual sentences are rarely simple sentences, and in most cases they are phrases? It is a complex sentence containing multiple units.

In the case of a simple sentence, all noun clauses preceding the predicate are related to the predicate, so there is no difference in dependencies.

Therefore, the analysis at this time is mainly based on the case and meaning of the predicate of each noun clause.

However, in the case of complex sentences, there is generally ambiguity in the dependency between the noun clause and the predicate.

Conventionally, Japanese grammatical information other than the case-dominant Senzo, or the typical compound sentence ': Turnke, which is prepared in advance, has been used to make use of Aisyisa? efforts are being made to reduce it. Although this may partially solve the problem, the general problem is that ambiguities are not resolved or the analysis is often incorrect.

[Purpose of the invention]

The purpose of the present invention is to solve the problems of dependencies in complex sentences that cannot be solved by the above-mentioned grammatical information or empirical complex sentence patterns. The objective is to provide a Japanese language analysis method that can be used at the analysis level to accurately analyze Japanese sentences.

[Summary of the invention]

In order to achieve the above object, the present invention provides a word dictionary that stores Japanese words and their related grammatical and semantic information, a word division unit that uses this to divide a Japanese sentence into word elements, and a phrase dictionary. Grammatical relationships between word elements within? The stored phrase-like grammatical information storage section, the phrase assembly section that assembles phrase 2 from the divided word elements and outputs the sounds of the phrase sequence, and information regarding the grammatical and semantic relationships between the phrasal and the word it controls. ? In the Japanese language analysis method that includes a dependency analysis unit that uses the stored case pattern dictionary to perform all the dependency analysis of Japanese consisting of a string of phrases, it is possible to match the case particles and meanings of phrases and case patterns. judgement? By using this method, you can determine the dependencies of a clause, and if all the cases in the case pattern of the disjunction have been matched, and there is a clause that does not have a case to match, use this, That phrase? It is characterized by the dependency analysis of later predicates.

[Embodiments of the invention]

The following is a drawing of an embodiment of the present invention. This will be explained in detail using

FIG. 1 is a block diagram showing an embodiment of a Japanese language analysis system for realizing the present invention. l is word segmentation part, 2
3 is a clause assembly section, 3 is a dependency analysis section, 4 is a word dictionary storage section, 5 is a clause-like grammar information storage section, and 6 is a case pattern dictionary storage section.

An input sentence 200 created using a keyboard such as an electronic lifter is input to a word dividing section l.

The word dividing unit l uses the word information stored in advance in the word dictionary storage unit 4 to divide the input sentence into a string of 200 pieces of candy (
It is divided into word strings) and sent to the next clause assembly section 2. The word dictionary storage unit 4 stores Japanese words and grammatical information, semantic information, etc. of the words, and when the input sentence 200 is divided into words by the word division unit l, the words attached to the words are stored. The grammatical information and semantic information are sent to the next phrase assembly section 2 together with the word string data 210.

Based on the received word string data 210, the phrase assembling section 2 uses the phrase white sentence it*@i stored in the phrase-based grammar information storage section 5 to assemble a plurality of phrase units into the word string data 210. Assemble to. Normally, a Japanese phrase is considered to consist of one independent word followed by zero or more dependent candy, and the phrase-based grammar information storage unit 5 stores the information of these independent words and attached maple candy. Connection conditions are memorized. Additionally, in general, the type of modification that indicates whether the clause is related to the following nominal or predicate can be determined by the candy at the end of the clause, and information indicating this is also stored in the clause-based grammatical information storage unit 5. ing.

For example, in the word string data 210, "...I am..."
When we write the word string, ``wa'' is an independent word, and ``wa'' is a dependent particle, which is one of the attached & words. The dependent particle is
Normally, conjunctive modifier clauses (clauses related to predicates) are formed together with nouns, so the clause assembling unit 2 uses this information to cut out "Watashi wa" purple from the word string data 210 and assemble it into a clause. And this is determined to be a conjunctive modifier clause.

The dependency analysis unit 3 uses the case pattern information of the predicates stored in the case pattern dictionary storage unit 6 to analyze the 220 clause strings sent from the clause assembly unit 2, and determines the dependencies of the clauses. 230 analysis results are output.

FIG. 2 shows the internal configuration of the dependency analysis section 3.

Reference numeral 31 denotes a phrase string storage section, in which the phrase string data 220 sent from the phrase assembling section 2 is temporarily stored. Reference numeral 32 denotes a simple sentence extraction section which scans the clause string storage section 31, cuts out single sentence units consisting of one predicate and one or more conjunctive modifier clauses, and extracts them into the clause string storage section 34 and the conjunctive modification clause storage section 35. The obtained simple sentence data is stored and passed to the control case pattern matching unit 33.

The case pattern matching unit 33 has a conjunctive modifier clause storage unit 34,
Following the data stored in the predicate phrase storage section 35, first, based on the contents of the predicate phrase storage section 35, refer to the case pattern dictionary storage section 6 to determine the case pattern of the predicate of interest,
The case pattern is imported into the temporary storage section 36. Next, using this case pattern, all lines of the dependency analysis of the phrase in the conjunctive modifier phrase storage section 34 are written into the corresponding slots of the resultant phrase string storage s31. The analysis is performed by the simple sentence extraction section 32.
The phrases are performed one by one in the order starting from the phrase most recently registered in the conjunctive modification phrase storage section 34. After completing the processing of all the conjunctive modifier clauses that can be analyzed, the case pattern matching unit 33 notifies the simple sentence extraction unit 32 that the processing has been completed. At this time, the dependency analysis may fail and some phrases whose dependencies cannot be fixed may remain, but these will remain in the conjunctive modifying phrase storage section 34, and those for which the analysis is successful will be deleted from the conjunctive modifying phrase storage section 34. do.

When the simple sentence extraction section 32 receives the completion report from the case pattern matching section 33, the simple sentence extraction section 32 selects the phrase string remaining in the phrase string storage section 31. Scan and single sentence unit? It is cut out and stored in the conjunctive modifier phrase storage section 34 and the pragmatic phrase storage section 35, and then the case pattern matching section 33 is activated again. At this time, it can be seen that there are phrases left as a result of the previous analysis failure in the conjunctive-modifying phrase storage section 34, but in this case, the continuation-modifying phrase storage unit 34 may contain phrases that have been left as a result of failure in the previous analysis. Write. However, the case pattern matching unit 33 performs all analyzes without distinguishing between these remaining clauses and newly cut clauses.

The cycle consisting of the above-mentioned simple sentence extraction s32 and the alternating processing by the case pattern matching unit 33, the extra-bunsetsu storage unit 31
This continues until there are no more phrases to be analyzed, and as a result, nine results are obtained in the phrase string storage section 31. It is output as an analysis result 230.

FIGS. 3 and 4 are flowcharts each showing an example of the operation of the simple sentence segmentation section 32 and case pattern matching g33.

5 shows an example of a phrase string generated in the extra phrase storage section 31 as a result of the processing by the word segmentation section 1 and the phrase assembly section 2. In FIG. 5, the column of dependent phrases and the column of case are blank, but necessary information is filled in after processing by the case pattern matching section 33.

3rd cause ~ 5th figure? The operation sounds of this embodiment will be explained in more detail using examples.

(A) Processing in the simple sentence extraction section 32 (Part 1) First, the reading position of the phrase string storage section 31 is adjusted to the beginning of the entire phrase string storage section 31 (step 101). Next, it is checked whether all the data in the bunsetsu string memory section 31 have been read (step 102), but since this is the first time, proceed to the next step and read the bunsetsu column storage section 31 to find the bunsetsu string %? Read out one r item (step 103), and advance one position to the reading position in the phrase string storage section 31 (step 104). The clause just read out is ``I am'' as shown by the reference numeral 231 in Figure 5.

This clause (because it is not a 1-phrase clause (step 105)
, this clause? Step 106) returns to step 102 to register it in the continuous modifier phrase storage m34. Then, the process proceeds to steps 102, 103, and 104, and then "sakura ga" indicated by the reference numeral 231 is called, but since this is also not a pragmatic clause, it is stored in the conjunctive modifier clause storage unit 34, and step 102
Return to step 105, 106. Further, when the process proceeds to steps 102, 103, and 104, the phrase ``Sakumono de'' indicated by the reference numeral 231 is read out, but since this is a pragmatic phrase, the process proceeds to step 107, and (step 105)
, the first phrase is registered in the phrase storage unit 35, and the case pattern inquiry unit 33? Start (steps 107, 108
). In this state, the predicate clause storage unit 35 and the conjunctive modifier clause storage unit 340 are shown at 235 and 234 in FIG.

Next, the operation of the case pattern matching section 33 will be explained with reference to the flowchart shown in FIG.

(B) Processing of the case pattern matching section 33 (Part 1) When the case pattern matching section 33 is activated by the simple sentence segmentation section 32, the case pattern matching section 33 first stores a case pattern dictionary based on the contents of the pragmatic clause storage section 35. The corresponding case pattern is read out from the memory fft16 and stored in the case pattern temporary storage section 36 (step 109).

What is currently stored in the pragmatic phrase storage unit 35 is "Sakumaru wo" as indicated by the reference numeral 235 in FIG. 6, and since its basic form is "Saku", 1 The case pattern of ``Saku'' is read out and stored in the case pattern temporary storage section 36.

The contents of the case pattern temporary storage section 36 at this time are shown at 236 in FIG. Here, to simplify the explanation, we assume that there are three cases of ``bloom'': ``living body,''``time,'' and ``place.'' Corresponding to each case, the particle expressing the case and the meaning information of the word standing in that case are read out from the case pattern dictionary storage section 6 and stored in the case pattern temporary storage section 36. Ru. The collated flag column indicated by reference numeral 236 in FIG. 7 is used to determine the case collation status at the time of processing, and is initially blank.

Next, in step 110, it is checked whether there are any remaining clauses to be processed in the conjunctive modifier clause storage unit 34, but since this is the first clause, the process proceeds to step 111, and from the end of the conjunctive modifier clause storage unit 34, the clauses to be processed are checked. Read out one data box. In this case, as can be seen from the reference numeral 234 in FIG.
The cherry blossoms are.

Next, in step 112, case pattern-1 temporal storage unit 3
It is checked whether there are any unmatched cases remaining in 6, but since this is also the first case, the process goes directly to step 113 and compares the particle of the clause, ``ga'' of ``sakuraga'', with the unmatched case of the case pattern. Match r with the particle in the case slot. in this case,
Looking at reference numeral 236 in FIG. 7, since the particle "ga" of the subject case matches, (12) the check in step 114 is successful and the process proceeds to step 115. In step 115, the meaning of the case indicated by the particle is compared with the meaning of the independent word of the conjunctive modifier clause. Now, the independent word of the bunsetsu is "cherry blossoms" and the meanings in the case of "subject" in the case pattern are "plant" and "flower", so the meanings match, so the check in step 116 is successful. , proceed to step 117. Note that although the semantic information of independent words is not clearly shown in Figures 5 and 6, the word segmentation f in Figure 1
During processing in l11, the phrase data 2 is extracted from the word dictionary storage section 4 along with grammatical information, etc., and stored in the phrase string storage section 31.
It is passed together when 20 is written.

In this case, semantic classification data such as "plant" for "cherry blossoms" is passed as semantic information.

In step 117, as a result of matching between the phrase and case particles and meanings, the dependency and case information is written into the corresponding position of the phrase string storage unit 31 as indicated by reference numeral 331 in FIG. 8. In this case, "cherry blossoms" means "because they bloomed."
Therefore, as dependency information, the clause number 3 of the dependent clause "Sakura ga Ko's corresponding position (13) is honour." The name of the case, in this case the ``subject'', is the name of the case. Further, as a sign that this case has been used, 'l'' is written as a verified flag in the case pattern temporary storage unit 36, that is, in the verified flag field indicated by the 7th symbol 236 (\J).

In step 118, "Sakura ga" whose dependency analysis was successful is
delete the clause data from the conjunctive modification clause storage unit 34,
Return to step 110.

Bed degree, step 110. However, this time, since "Sakura ga" has been deleted from the conjunctive modifier phrase storage section 34,
As shown by reference numeral 334 in FIG. 9, the last word is "I am", which is read out. Next, step 112,
113, the particle of the clause is compared with the particle of the case pattern, but this time the particle of the clause is ``wa'', and there is no match with this in the case pattern shown by reference numeral 236 in FIG.

However, in the case of Japanese, from the point of view of case grammar, when ``wa'' appears alone, it can be replaced with ``ga'' or (to).

(14) This point is illustrated on page 174 of paper 1, “Description of Japanese Generated Words and its Application to Pregnancy Processing,” published in “National Insurance Research on Electronic Computers ■” published by the National Institute for Japanese Language and Linguistics. has been done.

According to this, when we replace ``ha'' ke, ``ga'' ka, or ``wo'' in the case pattern matching section shown by reference numeral 236 in Figure 7, there is no corresponding one for ``ge''. You can see that there are matches for "ga".

However, in the case corresponding to ``ga'', that is, the case ``subject'', 'l'' was set in the matching technique flag in the previous process, and the matching has already been completed.In general, two or more different clauses do not stand in the same case (easily 109, for example, one phrase does not have two main chains?), so the case for ``ga'' and the case for ``subject'' also have the clause `` I can't match 'ke'. Therefore, as a result of step 113, the determination process of step 114 fails, and the process moves to the end process of step 119, where the case pattern matching unit 33 sends a completion report to the simple sentence extraction unit 32, and the process ends.

(15) At this stage, the phrase string storage unit 31. The contents of the conjunctive modifier clause storage section 34 and the case pattern-poetic storage section 36 are indicated by reference numeral 3311 in FIG. 8, reference numeral 334 in FIG. 9, and reference numeral 336 in FIG. 1O.

(C) Processing of simple sentence extraction 111132 (part 2) When the simple sentence extraction unit 32 receives a processing completion report from the case pattern matching unit 33 (step 100 in FIG. 3), it again searches the phrase string storage unit 31? Scan and cut out a new simple sentence. However, in the previous process, the reading of the phrase data stored in the phrase string storage unit 31 up to phrase number 3, "Because it bloomed," has been completed, and the next reading position is set to phrase number 4, "To the park." So read from there? Start.

The phrase data “to the park” and “went to the park” are read out in the same process as the previous time, and are stored in the conjunctive modifier phrase storage unit 34.
The words are registered in the phrase storage section 35 for use in the work. However, this time, the clause data ``Watashi wa'' is stored in the conjunctive modifier clause storage unit 34 as a leftover from the previous case pattern matching unit 33.
remains, and (16) the phrase data ``to the park'' is registered behind it, as indicated by reference numeral 434 in FIG. 11.

As a result of the above processing, the conjunctive modifier phrase storage unit 3 at this stage
4 Contents of the phrase storage section 35 Reference numeral 4 in Fig. 11
35 and 434.

Next, the process proceeds to the case pattern matching section 33 again.

(D) Processing of the case pattern matching section 33 (Part 2) As in the previous case, the case pattern matching section 33 first refers to the case pattern dictionary storage section 6 and reads out the case pattern corresponding to the predicate. Now, the predicate phrase is "go" and its basic form is "go", so the case pattern corresponding to "go" is read out and stored in the case pattern temporary storage section 36. At this time, the contents of the case pattern temporary storage unit 36 Kintei 1
It is shown at 436 in FIG.

As in the previous case, the case pattern matching unit 33 references the contents of the case pattern temporary storage unit 36 (reference numeral 436 in FIG. 12) and matches the clauses and cases one by one starting from the back of the conjunctive modifier clause storage s34. Let's do it.

(17) First, referring to the contents of the case pattern temporary storage unit 36, "to the park" with the reference numeral 436 in FIG. Since they are a type of ``place'', they match, so ``to the park'' is a case of ``place'' and is related to ``went''. This information is loaded into the phrase string storage section 31.

Next, the phrase data "I am" is processed. As mentioned above, the particle "wa" can be replaced with "ga" or "wo", and the contents of the case pattern temporary storage unit 36 (reference numeral 436 in FIG. 12) k If you refer to '
There is a match in the case of ``subject,'' and in this case, since ``I'' is a type of ``person,'' the meanings also match, and the match between the clause and the case is successful. This data is written into the phrase string storage section 31.

With this, the processing of the phrase data in the conjunctive modifier phrase memory 11fl134 is completed, and the case pattern matching unit 33 proceeds to the end process based on the determination in step 100 in FIG. 4, and reports the completion of the process to the simple sentence extraction unit 32. 18) The processing is finished.

(E) Processing of simple sentence extraction s32 (Part 3) When the simple sentence extraction unit 32 receives the completion report (step 10 in FIG.
0), the contents of the extra-clause storage unit 31 are scanned six times, but since all the contents of the extra-clause storage unit 31 have already been read, the determination in step 102 is directed to completion, and the process is completed.

With this, the processing of the dependency analysis unit 3 is completed.

As a result of dependency analysis, extra-clause memory section 3 finally obtained
Contents of 1? It is shown at 431 in FIG.

This becomes the analysis output 230 of the dependency analysis unit 3.

In this example, a sentence containing 42 predicates was used, but sentences containing one or three or more predicates i can be analyzed using the same procedure.

The detailed operation of this example is explained above using the example problem. As I explained, is the example problem used here about particle agreement and disagreement? For OU,
It was to eliminate one problem in the relationship (it was determined that ``Watashi wa'' was not related to ``Hasamagaro'').

Next, there are some particles that match, but the meaning is - (19) Does it not match? The second step is to perform all the dependency analysis using
An example problem will be explained.

Reference numeral 531 in FIG. 14 is a second example consisting of sentences generated in the extra-clause storage unit 31. In this example, the simple sentence extraction section 32 scans the extra-clause storage section 31r and cuts out simple sentences, and the conjunctive modification phrase storage section 34 and the phrasal phrase storage section 3
5 stores the phrase data indicated by reference numerals 535 and 534VC in FIG.

Next, the case pattern matching unit 33 reads out the case pattern t1 corresponding to the contents of the phrasal phrase storage unit 31, that is, the basic form “Mawaru” of “Mawaru”, from the case pattern dictionary storage unit 6,
The case pattern is stored in the temporary storage unit 36 (numeral 536 in FIG. 16). Next, the conjunctive modifier clause storage unit 34 reads the clauses one by one from the bottom and performs all clause and case matching. First of all, it is ``toride'', but there is a case corresponding to the particle ``ke'' that indicates the object of the action. Next, if we look at the meaning part of the target case, it says "It's an object," and since "Toride" is a type of object, the semantic matching is successful, and (20) "Toride" can be considered as the target case. It can be seen that this is related to

Next is ``Door wa'', but ``HA'' can be read as ``GA'' or ``AI'' as mentioned above.

The case pattern indicated by reference numeral 536 in FIG. 16 includes both "ga" and "l"k. However, since the case for ``1 ke'' is ``toride 葡'', which is already in use, the case for ``ga'' is the subject of matching. [If you look at the meaning of the case for J, that is, the case of the subject, there is one person.'' On the other hand, the clause is “
However, since "door" is not "enter", the meaning verification fails, and as a result, the determination process at step 116 in FIG. 4 moves toward the end process. Therefore, "Door wa" is left in the conjunctive modifier phrase storage unit 34 as unprocessed, and is subjected to processing of the next simple sentence.

In the next cycle, "open" is cut out as a phrase for the simple sentence cutout section 32Vi. There is no target to be extracted as a conjunctive modifier clause. Therefore, the phrase to be compared with the case frame of the basic form of "open" is the one remaining in the conjunctive modifier phrase storage unit 34: 1) "Door wa." The case frame for ``open'' is not shown here, but as can be easily seen, the clause ``door wa'' is a subject case, ``related to open.

As a result of the above processing, the analysis output 230 is the number 6 in FIG.
31.

Furthermore, if all the cases listed in the case pattern of a predicate have been matched with clauses in a simple sentence, but there are still unprocessed clauses, the case to be matched for that clause is Is that clause using the fact that it doesn't exist? The third method correctly parses the dependency by passing it on to the processing of subsequent predicates.
An example of this is shown below.

Reference numeral 731 in FIG. 18 is a third example problem consisting of sentences stored in the extra-clause storage unit 31. In contrast, first, the simple sentence cutting unit 32 selects the nineteenth code 735 and the code 734.
A simple sentence element as shown in FIG. The basic form of the pragmatic phrase ``go, so'' is ``go'', and its case pattern is the 12th case pattern.
It is shown at 436 in the figure. Using this sound, case pattern matching s33 matches all lines of phrase data and cases (22
) cormorant. There are three cases in the case pattern of ``go'', and the continuous modifier clause memory i34 has all the cases corresponding to these three cases. As a result of these collation processes, the contents of the extra-clause storage section 31 become as shown by reference numeral 831 in FIG. In this state, the phrase ``he'' remains unmatched, but as mentioned earlier, the cases in the case pattern of ``go'' are all matched with other phrases, and the phrase ``he'' remains unmatched. ``r'' There is no case to check. Therefore, the determination process at step 112 in the flowchart of FIG. 4 proceeds to the end process. As a result, since there is no case to match, the phrase ``he'' can be used to process the subsequent pragmatic phrase. In the next cycle, ``he'' is matched with the case pattern of the basic form ``help'' of the predicate ``helped'' along with ``packing rikke''. This process is similar to the previous example, and although it will not be described in detail here, it will be correctly analyzed.

As described above, the dependency analysis of the phrase string shown in FIG. 18 is performed, and finally an analysis output 230 shown at 931 in FIG. 21 is obtained.

(23) In this example, the term ? Although we have explained an example of a sentence containing two predicates (f-3 or more), it is also possible to analyze sentences containing three or more predicates using the same procedure.It is also possible to analyze sentences containing only one predicate. It is clear that

Furthermore, it goes without saying that if the analysis method described in this embodiment is used together with grammatical information other than the case-dominant structure of the predicate, it is possible to perform the entire analysis with high accuracy.

〔Effect of the invention〕

As is clear from the above explanation, according to the present invention, case pattern information of predicates? By using this method, you can perform all analyzes of complex sentences, and because all analyzes are performed using meanings, you can achieve higher-precision analysis.

[Brief explanation of the drawing]

FIG. 1 is an overall configuration diagram of a Japanese language analysis device showing one embodiment of the present invention, FIG. 2 is a configuration diagram of dependency analysis, FIG. 3 is a flow chart showing an example of the operation of the simple sentence extraction section, A) and (B) are the operations of the case pattern matching section (24). Flowcharts showing examples, FIG. 5, FIG. 6, FIG. 7, FIG. 8, FIG. 9, FIG. 1O, FIG. Figure 12, 13
14, 15, 16, and 17 are explanatory diagrams of operation examples of the dependency analysis unit according to the second example, Figure 18,
FIG. 19, FIG. 20, and FIG. 21 are explanatory diagrams of an example of the operation of the dependency analysis unit according to the third example problem. l...Word division part, 2...Phrase assembly part, 3...
- Dependency analysis section, 4... Word dictionary storage section, 5...
Intra-clause grammatical information storage unit, 6... Case pattern dictionary storage unit, 31... Out-of-clause memory unit, 32... Simple sentence extraction unit, 33... Case pattern matching unit, 34... Conjunctive modification clause Memory section, 35... Dative phrase memory section, 36... Case pattern temporary (25) No. l Country 30 No. ? Circle 3 5th eye 4th day 4th mouth (b) 6th figure 7th day 8th day ? Country No. 10 /l Country No. 12 Figure 74 / Furukuchi No. 76 ■ No. t7'h 6vt Figure 18 JP-A Showa GO-157659 (1 ■

Claims (1)

[Claims] 1. A word dictionary storage unit that stores all Japanese words and related grammatical information and semantic information; a word division unit that uses this to divide Japanese words into Japanese word elements; Single V&
An intra-clause grammatical information storage unit that stores grammatical relationships between elements, a clause assembly unit that assembles clauses from the divided word elements and outputs a clause string, and a predicate and a predicate that governs the clause. A case pattern dictionary storage unit that stores all information regarding grammar and semantic relationships with words, and a case pattern dictionary storage unit that uses the information in the case pattern dictionary storage unit to perform uke analysis of a Japanese sentence consisting of a string of phrases. With the dependency analysis section? In the Japanese language analysis method that has a Japanese language analysis method, is it possible to determine the attachment of a clause by determining the agreement between the clause and the case particle of the case pattern and the meaning? Japanese language analysis method. 2. Determine whether or not there is an unmatched case remaining in the case pattern of the predicate, and if there is no unmatched case, change the unprocessed clause to the predicate after the currently focused predicate. 1. A Japanese language analysis method according to claim 1, wherein the Japanese language analysis method is characterized in that dependency analysis processing is performed on the following.