JPH04174070A - Natural language sentence analyzing device - Google Patents

Abstract

PURPOSE:To eliminate the need for the knowledge of an object language by extracting one or plural pieces of information for discriminating between plural functional structures and other functional structures by an interactive process part which performs various-meaning dissolution process after a syntax analysis by a syntax analytic part, and performing the extraction on an interactive basis. CONSTITUTION:An input part 1 inputs information from an operator at the time of the interactive various-meaning dissolution process. The syntax analytic part 3 divides an input sentence from an input part 1 into morphemes and plural constituent structures are generated for the divided morphemes; and functional structures are formed of the constituent structures and restrictions are applied when the functional structures are formed. After the syntax analysis by the syntax analytic part 3, an interactive process part 4 which performs the interactive various-meaning dissolution process extracts one or plural pieces of information for discriminating between plural functional structures and other functional structures and the extraction of one or plural pieces of information is performed on an interactive basis. Consequently, the knowledge of the language is unnecessary.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a natural language sentence analysis device, and more particularly to a natural language sentence parsing device used in realizing a system that receives natural language sentences as input, such as a machine translation device. This invention relates to an interactive syntax analysis device that uses dialogue with the user as an information source.

There are many sentences in natural language whose meanings are ambiguous. For example, (,) I read the book I bought yesterday.

It is unclear whether it means "I bought it yesterday" or "I read it yesterday." This ambiguity is ``required''
It is called ambiguity. In other words, the meaning of ``yesterday'' is ambiguous. Also, (b) I eat fish.

It is unclear whether r fish eats something j or ``something eats fish.'' This ambiguity is called ``role'' ambiguity. In other words, "eat" in "fish"
This means that it is ambiguous whether the role of the word is the subject or the object.

For humans, these ambiguities in natural language are
In speech, it is thought that the situation of the utterance, intonation, pauses, etc. are resolved as information.
In a syntax analysis device used for machine translation, etc., only character string data input through an input unit is considered as input data, so it is difficult to eliminate the above-mentioned ambiguity. Therefore, many proposals have been made for techniques in which the operator intervenes to obtain the results intended by the operator. Regarding machine translation processing, which is an application field of natural language sentence analysis devices, for example, Japanese Patent Application Laid-Open No. 63-300360
No. 59-140582, Japanese Patent Application Laid-open No. 1982-140582
-18073, etc. Hereinafter, the technology for narrowing down the analysis results of ambiguous sentences will be referred to as the technology for "resolving ambiguity."The technology described in Japanese Patent Application Laid-Open No. 63-300360 is designed to improve the efficiency of translation and editing work. The editing control unit has the function of retrieving translated sentences that have already been translated from the translation storage unit and displaying them on the display unit while the translation is continuing in the translation unit, and in response to editing command information issued while the translation is continuing. This type of conventional technology and similar technologies are called post-editing methods, in which natural language sentence analysis devices force one of the possible analysis results. The translation generation unit of the machine translation device converts the unique solution into the target language. Correcting errors in translation results due to incorrect selection of an analysis device is performed by editing the target language itself.

In addition, what is described in Japanese Patent Application Laid-open No. 59-140582 classifies cases in which ambiguity occurs in natural language analysis in advance, and identifies input languages that belong to these classifications and are likely to cause ambiguity. Translation processing is performed by inserting instruction data into the relevant part in advance according to an instruction method that eliminates ambiguity before syntax analysis processing. This method eliminates ambiguity by adding information to resolve the ambiguity to the input sentence input to the natural language sentence analysis device.

Furthermore, in the method described in Japanese Patent Application Laid-open No. 61-18073, proofreading is performed at the end of the analysis of the first language in the translation process, and, for example, the analysis results are displayed as an analysis tree and the node numbers are used. This method performs proofreading and uses macro commands for a series of frequently used proofreadings.This conventional technology and similar technologies are called interedit methods, in which a natural language sentence analysis device analyzes input sentences. The analysis result is shown to the operator, and if the analysis result is incorrect, the operator corrects it himself, and the corrected result becomes the object of conversion generation processing.

In addition, for interactive natural language sentence analysis devices that attempt to eliminate ambiguity through dialogue with the operator, there are examples such as ``On a Method of Dialogue Translation'' (4 people from Aohaku, Technical Research Report of the Institute of Electrical Information and Communication Engineers). NLC90-14, P, 17-24.
There is a method described in (1990). These methods are called interactive methods. In the prior art described above, the operator resolves the ambiguity in an editing format, but in the prior art described in this document, when the system detects an ambiguity, the information necessary to resolve the ambiguity is generated. It is an interaction format in which ambiguity is resolved by asking the operator and the operator responding.

On the other hand, when parsing a natural language sentence, it is difficult to perform precise analysis just by constructing a structural data structure such as a parse tree. Therefore, in addition to the structural data structure, a semantic data structure is constructed, and as a grammar theory that provides a framework for describing grammar complementary to these two data structures, the Mental Representat
ion of GrammaticalRelation
ns, J (Bresnan, J+A IT Press
, 1982) discloses an overhead functional grammar.

The problem with the post-editing method is that knowledge of the target language is required in order to edit the target language. For example, if a Japanese person wants to use a Japanese-English machine translation device, knowledge of English is also required.

Furthermore, although the pre-editing method does not require knowledge of the target language, it is necessary to anticipate the ambiguity of the input sentence. In order to add enough information to resolve ambiguity, it is necessary to add a lot of information to various parts of the input sentence, and for this to be necessary and sufficient, sufficient knowledge of the language and system processing are required. requires sufficient knowledge of In other words, it is difficult to add the minimum amount of information necessary to obtain the analysis results intended by the operator.
This inevitably adds redundant information, which is wasteful. Further, the work itself of adding information to each input sentence is complicated.

In addition, regarding the inter-editor method, a tree structure or similar structure is displayed as the analysis structure of the input sentence.
The operator corrects and confirms the display. However, there is a problem that the operator cannot understand even if the complicated structure of the input sentence is shown to him.

In terms of the prior art described above, the present invention belongs to the interactive method. However, in conventional interactive systems, a gap between the information required by the system and the information that the operator can respond to has become a problem.

For example, questions about "dependency" or "roles" are fine, but questions about the meaning of polysemous words are often difficult for operators without sufficient grammatical knowledge. For example, 1 in Japanese: "The book I bought was..."
Even when asked whether the meaning of ``reru'' in the sentence ``reru'' is ``passive'' or ``respect,'' an operator who is not familiar with Japanese grammar would not be able to understand the meaning. Therefore, the answer will probably be arbitrary. Conventional interactive parsers treat the information obtained through dialogue with the operator as absolute and reject solutions that do not match the operator's answer, so if the operator gives a wrong answer, , it becomes impossible to obtain a solution, and the questions from the system themselves do not take into account the grammatical knowledge of the operator.

The present invention was made in view of the above-mentioned circumstances, and
When the natural language sentence analysis device detects ambiguity in the input sentence when analyzing the input sentence, it asks for enough information to resolve the ambiguity at that point in the most efficient order, and this question Since the questions are "receive" or "role" questions, knowledge of the target language is not required, and the questions are asked when ambiguity is detected, and the questions are asked in the most efficient order, so the questions are ambiguous. In addition, the content of the questions is about dependencies in the input sentence and the role of syntactic elements, and the interface that is difficult for the operator to understand, such as showing the parsed structure, is eliminated. In addition, for each question we have information about the simplicity of that question, in other words, the certainty of the operator's answer to that question, and we use that information to calculate efficiency and By handling answers to questions with high certainty and questions with low certainty differently, we do not require more grammatical knowledge from the operator than necessary, and we can also answer questions that require a lot of grammatical knowledge. was developed with the aim of providing a natural language sentence analysis device that does not allow incorrect answers.

1-1 In order to achieve the above objects, the present invention provides (1) an input section that inputs natural language sentences, morpheme segmentation of the input sentence from the input section, and morpheme segmentation of the morpheme segmented sentences. A syntax analysis unit that creates a plurality of constituent structures, creates a functional structure from each of the constituent structures, and applies constraints when creating the functional structure, and after the syntax analysis by the syntax analysis unit,
It consists of a dialogue processing unit that performs ambiguity resolution processing through dialogue, and an output unit that outputs the analysis result by the dialogue processing unit,
The interaction processing unit extracts one or more pieces of information that distinguishes it from other functional structures from each of the plurality of functional structures, and performs one or more of the pieces of information in an interactive format, or (
2) An input section that receives a natural language sentence as input, and divides the input sentence from the input section into morphemes, creates multiple component structures for the morpheme division, and generates a component structure from each of the component structures. A syntax analysis unit that creates a functional structure and applies constraints when creating the functional structure; and a syntax analysis unit that extracts one or more features that distinguish each of the multiple functional structures from other functional structures, and It consists of a dialogue processing unit that asks one or more of the questions and uses the answers to perform ambiguity resolution processing, and an output unit that outputs the analysis results of the dialogue processing unit, and the dialogue processing unit Having information regarding certainty; and (3) in (1) or (2) above,
The syntax analysis unit has a calculation means for calculating a value related to incompatibility with each functional structure, and the interaction processing unit calculates a value related to incompatibility with each functional structure. The feature is that dialogue is conducted based on values related to non-conformity.

In the natural language sentence analysis device according to the present invention, grammatical properties are expressed as constraints, and information regarding the grammatical strength (few exceptions) of each constraint is added to each constraint. On the other hand, a structure corresponding to a functional structure in a known functional grammar is constructed for the input sentence, and several constraints for realizing the functional structure are checked. Since each constraint has information on the number of exceptions, the information on the number of exceptions mentioned above is comprehensively evaluated for constraints that are not satisfied, the evaluation is expressed numerically, and it is regarded as an incompatibility of the functional structure. This nonconformity is called a penalty.

When an ambiguous sentence is input, a plurality of functional structures are obtained by calculating penalties for each. The penalty expresses the low probability that the functional structure is the correct analysis result. In the present invention, a structure representation vector is constructed by extracting information sufficient to distinguish the functional structures from other analysis results from each of the plurality of functional structures. Here, the structure expression vectors extracted from each functional structure sufficiently express each functional structure. The operator is asked the most efficient question for specifying one vector from the plurality of penalized structure expression vectors thus obtained. At this time, values related to the simplicity of the question and the certainty of the answer are calculated for each question, so these values are also taken into consideration when calculating efficiency. The most efficient question to identify one vector from multiple structural representation vectors is to identify one functional structure from multiple functional structures because the structural representation vector sufficiently represents the functional structure. However, this is the most efficient question after considering the operator's grammatical knowledge. Furthermore, regarding the operator's answer, by using the certainty of the answer as a penalty for functional structures that do not match the answer, processing that reflects the operator's grammatical knowledge becomes possible. Hereinafter, the present invention will be explained based on examples.

FIG. 1 is a block diagram for explaining one embodiment of a natural language sentence analysis device according to the present invention, in which 1 is an input section, 2 is an output section, 3 is a syntactic analysis section, and 4 is a dialogue processing section. , 5 is the dictionary section,
6 is a syntax rule section, 7 is a constraint section, and 8 is a function relationship section.

The input unit 1 includes input from an input device such as a keyboard or a storage device. The input unit 1 also serves as a means for inputting information from the operator during ambiguity resolution processing through dialogue.

The output unit 2 is a display or the like for conveying the analysis results or the content of the question to the operator, and also includes cases where an application using a syntax analysis device attempts to use the analysis results, such as in the case of machine translation.

FIGS. 2(a) to (c) show a dictionary D, which is part of the contents of the dictionary part, and a constraint C, which is part of the contents of the syntactic rule R1 constraint part, which is part of the contents of the syntax rule part. FIG. 4 shows part of the contents of the function-related section.

Each element of the dictionary shown in FIG. 2(a) is expressed by a headword, a category (part of speech), and a feature list. The feature list is a list of features, and each feature is a pair of the form (feature name, feature value). Here, the characteristics are extracted characteristics of miscellaneous expenses.

The syntactic rule R shown in FIG. 2(b) is written in a labeled phrase structure grammar that is an extension of the notation of a known phrase structure grammar. Each element on the right-hand side is either a labeled nonterminal or an unlabeled nonterminal.

In the case of (N P ; case) in R1, NP is a non-terminal symbol and the label case is added. Also, the non-terminal symbols on the right side, lowercase letters, are pre-terminal symbols (language range 1ll).
), uppercase alphabetic characters are nonterminal symbols other than preterminal symbols. Also, the labels in the syntax rules represent function names. Here, the term "function name" is used in the same manner as in the well-known function grammar. That is, the description of R1 is the same as the following RIO in the word-function grammar.

RIOVP->NP VP ↓=↑case ↓:↑ Each element of the constraint C shown in FIG. 2(c) is a production rule with a function name and a penalty.

The notation is in the form of a (Function name: Penalty) constraint rule. Here, the penalty is a numerical value representing the grammatical strength of each constraint, and the larger the value, the stronger the grammatical constraint, that is, the fewer one exceptions. Each constraint rule references information in the functional structure, specifically features from the dictionary.

Here, the term "functional structure" is used in the same manner as in the well-known overhead functional grammar, and is a recursive matrix in which the functional name is the attribute name and the functional structure is the attribute value. FIG. 6 shows an example of the functional structure.

FIG. 3 is a flowchart of the syntactic analysis section and the dialogue processing section of the natural language sentence analysis device according to the present invention.

Hiroko Tate: The syntactic analysis unit receives a Japanese sentence written in “solid writing” from the input unit and divides the input sentence into a list of morphemes according to morphological analysis processing. For example, if the input sentence is (aO) He plays in the park, the result of morpheme segmentation is (al) (He: n) (Ga: P) (Park: n) (De: P)
(Play: v) It becomes. Here, each morpheme is expressed in the format (headword 2 miscellaneous expenses range II). In morpheme segmentation, a dictionary for morpheme segmentation is extracted from the dictionary section and used.

Next, a compositional structure is created using the contents of the syntax rule section for the morpheme sequence after the morphological analysis. Here, the compositional structure is used in the same way as in the well-known expense function grammar, and is a labeled tree structure as shown in FIG. Labels attached to the tree structure represent function names, and can be obtained by referring to the labels described in the syntax rules when applying the syntax rules.

Regarding the process of creating a compositional structure from a sequence of morphemes,
Although many methods have been proposed and are known, here, the CKY method is used from the bottom up. When creating a compositional structure from a morpheme sequence, when there are no more morphemes to process, all compositional structures that have reached the final state are processed in 5 steps.
2, and there are generally multiple solutions.

Here, whether the final state has been reached is determined by whether the root node of the compositional structure created for the entire sentence is a non-terminal symbol S.

One of the construction structures created from the morpheme sequence of (al) (
a2) is shown in FIG.

A functional structure is created from each of the plurality of constituent structures obtained in Step 2 above. This process may follow the process of creating a functional structure from a constituent structure in the well-known overhead functional grammar, or it may be possible to follow the process of creating a functional structure from a constituent structure in the known expense function grammar, or to write a child node without a label in the constituent structure as a main word in the entry section, and to write it to a labeled child node. , the functional structure may be created recursively by using the label as the attribute name and the functional structure of the child node as the attribute value. FIG. 6 shows a functional structure (C3) created from the component structure (C2) shown in FIG. 5. In this st'ep3, a functional structure action dictionary is extracted from the dictionary section and used. The contents of this dictionary are part of the feature list in the dictionary. Since there is a one-to-one correspondence between constituent structure and functional structure, there are multiple solutions for the functional structure in the output of 5tep 3.

Constraints are applied to the functional structure using the constraints from the constraint section and the constraint application dictionary from the dictionary section.

The procedure for applying constraints is as follows.

■For all functions in the functional structure, all constraints indexed by each function are applied, and for constraints that are not satisfied, the penalty of that constraint is added. At this time again,
Functional structures for which the penalty is sufficiently large are treated as analysis failures and no further analysis is performed.

■For a function to which constraints have been applied, refer to the function-related section and change it to a child function of that function. If there are multiple child functions at this time, the functional structure is copied to multiple functions.

■Repeat steps (2) and (2) above for the changed function name. Continue these ■, ■, and ■ until the function name cannot be changed.

The above steps (1) to (2) are performed recursively for the recursive structure of the functional structure. Here, each constraint has a function name, as shown in constraint C,
It is easy to refer to because it has a penalty attached to it. Furthermore, since the function names are managed in a tree structure as shown in FIG. 4 in the function-related section, it is easy to change each function to a child function.

An example of the processing in step 4 is executed for the functional structure (C3) shown in FIG. The function name corresponding to the phrase "in the park" is initially CaSe given by the syntax rule R1. Regarding this CaSe function, the constraint CO is applied according to the above-mentioned (2), but it is satisfied because the main term in the parent functional structure is the term "Mbu". Next, the case function is changed to its child functions 5ubj, obj, obj2, and obl according to the above item (2). In this case, since there are multiple child functions, the functional structure is copied to correspond to each child function. As an example, when the function is changed to 5ubj function, the constraints C1 to C4 are applied to the 5ubj function according to the above-mentioned (■). Among these, C3 is not satisfied because the value of the case marker feature in the 5ubj case is "de". Therefore, this functional structure is subject to a penalty of 200, but
This penalty is large enough that "in the park" is 5ub.
This functional structure, which has a j case, results in an analysis failure and no further analysis is performed. On the other hand, constraints C5 and C6 are applied to the functional structure in which the case function is changed to the ob1 function, but since both are satisfied, the penalty is O, and the function name is further changed in the above-mentioned ■. The above processing is performed in the clause ``
An example of the finally obtained functional structure is shown in FIG. 7. The penalty for this functional structure is O. The process in step 4 creates a plurality of functional structures for one functional structure. Since the input to 5tep 4 is a plurality of functional structures, the output of 5tep 4 is a plurality of functional structures as shown in FIG. 7, and a penalty has been calculated for each functional structure.

W: Select a functional structure based on the penalty.

Specifically, one or more functional structures with sufficiently low penalties are selected.

5t for unambiguous sentences like (a O)
There is only one functional structure selected in ep 5, as shown in FIG. 7 (penalty is on), so there is no need for ambiguity resolution processing through dialogue, and nothing is done in the next 5tep 6.

% ;Finally, the functional structure after the process of eliminating ambiguity through dialogue is output to the output section as an analysis result.

For the example sentence (al), the functional structure shown in FIG. 7 is output.

The above was about the unambiguous input sentence Δhe plays in the park. However, regarding (bo) I read the book I bought yesterday, as mentioned above, it is unclear whether ``yesterday'' refers to ``bought'' or ``read.'' FIG. 8 shows the structure of "yesterday" related to "bought".

In Figure 8, the function rentai#obj is the function renta.
It is a child function of i, and 1 function rentai is an adnominal modification clause (
(embedded sentence), the function rentai#obj means that the modified nominal matches the objective case in the modified lung. In this example, ``r book'' means that it is the objective case of ``bought''. FIG. 9 shows a functional structure (bl) created from the configuration structure shown in FIG. 8. Now, (bO) is "yesterday"
5 step 5 in Figure 3 because the person involved was ambiguous.
After processing, two functional structures shown in FIG. 7 are obtained, one shown in FIG. 9 and one (b2) shown in FIG. 10 (each penalty is O). Here, in FIG. 10, the features are omitted for simplicity. The difference between the two functional structures is
It can be seen that this is the position in the functional structure of Yesterday J.

FIG. 11 is a flowchart of the ambiguity resolution process by dialogue performed by the dialogue processing unit in step 5 of FIG. 3. Below, each step will be explained in order.

Phantom fLL''-; Ambiguity is resolved through dialogue when there are multiple functional structures. In that case, a structure representation vector is obtained for each functional structure.

For (bl), (b3) penalty = O (2 adverb 3 rentaistobj 4
For obj Ohead (b2), (b4) penalty = O (4adverb 3 rentai#obj 4 o
bj Ohead. Here, the odd-numbered elements of each vector represent the clause number of the decoration light of each clause in order from the front. For example, the first "2" in (b3) above indicates that the phrase "yesterday" modifies the second phrase "bought." Furthermore, rQJ indicates that there is no modifying morpheme. The even-numbered elements of each vector represent the role of each morpheme in the sentence in order from the front. Here, the "role" in the sentence is expressed by a function name. The second element in (b3) above, r adverb J, represents that the functional name of the morpheme "yesterday" is r adverb J. Here, each structure expression vector is sufficient to restore the functional structure using dictionary information, that is, it sufficiently represents the functional structure. In other words, differences in functional structure can be seen by comparing the structure expression vectors.

In the above examples (b3) and (b4), it is immediately clear that the difference is whether "yesterday" is related to "bought" or "read."

5. Since there are multiple functional structures, multiple structure expression vectors can be subdivided. From now on, we will narrow down this vector to one based on the information obtained through dialogue.

If the number is narrowed down to one, a functional structure corresponding to the structure expression vector is output in step 515.

"on n"-: If there are a plurality of structure expression vectors, find the sequence with the largest variation from the set of structure expression vectors. In the above example, the elements other than the first element are the same, so it can be seen that it is the first column.

Tangasa Kori: Ask the operator the sequence obtained in step 12 above. In the above example, what does (b5) r yesterday relate to? Or, the person in charge of (bO)r yesterday's "bought" and "read"
Which one? Just ask.

5. Leave only the structure expression vectors that match the answer to the question obtained in step 13 above, and treat other vectors as failures in analysis. The answer to question (b5) or (bO) is “I bought it.”
If so, only the structure expression vector (b4) is left.

Since the structure expression vector has been narrowed down to one at step 14, the functional structure (bl) corresponding to the structure expression vector is output.

Above, there is an ambiguity in ``I read the book I bought yesterday,'' but the ambiguity concerns the input sentence, which is one of the dependencies of ``yesterday.'' However, (e O) he read the book he bought yesterday.

Regarding ``yesterday,'' it is ambiguous whether it relates to ``bought'' or ``read,'' and it is also ambiguous whether ``he'' relates to ``bought'' or ``read.'' , there are two ambiguities. In other words, 5te in Figure 3
The parser in p5 outputs, for example, three functional structures. The structure expression vectors calculated by the interaction processing unit from each of them are, for example, the following three.

(el) Penalty = O (35ubj 3 adverb 4 rentaNf
obj 5 obj Ohead) (e2) Penalty = 10 (55ubj 3 adverb 4 rentai#
obj 5 obj Ohead) (e3) Penalty = 15 (55ubj 5 adverb 4 rentai#
obj 5 obj Ohead). What this result means is that (e4) r someone (probably the speaker) read the book he bought yesterday. ” is the most plausible meaning.

(e2) says, ``He read the book I bought yesterday.

'', (e3) corresponds to the interpretation ``Yesterday he read a book that (someone) bought.''

It can be seen that these two interpretations are less likely than interpretation (e4) based on the penalty values.

In this case, the dialogue processing section uses the penalty value of each structure expression vector to ask (e5) What is the role of "r"? (e6) What is the person responsible for "r yesterday"? Of the two possible questions,
Select (e5) and ask the operator. The reason is as follows.

The dialog processing unit predicts that the functional structure whose structure expression vector is (el) is the correct answer based on the penalty value.

Therefore, the operator's response is considered to be an answer that matches (el). Even if you asked about e6) first, you would probably get a response saying "I bought it." In that case, if you do not continue to ask question (e5), (e 1) (e
It is unclear which functional structure corresponding to 2) is correct. - question (e5) first, and if the answer is "I bought it," then it can be determined that the functional structure corresponding to (el) is the correct answer, regardless of (e6).

As described above, the dialogue processing unit performs dialogue in the most efficient manner based on the penalty added to each functional structure by the syntax analysis unit. Since this is nothing but a conversation that maximizes the amount of information, it can also be achieved by a method based on known information theory.

next, (fO) I read the book you bought yesterday.

Regarding ``yesterday'', it is ambiguous whether it relates to ``bought'' or ``read'';
It is ambiguous whether it means "passive" or "respect". Therefore, the syntactic analysis unit S5 has, for example, 3
It has two functional structures. The structure expression vectors calculated by the interaction processing unit from each of them are, for example, the following four.

(fl) Penalty = 0 (2adverb 3 rentaNIpass#ob
j 4 obj Ohead) (f2) Penalty =
10 (4adverb 3 rental#pass#ob
j 4 obj Ohead) (f3) Penalty =
0 (2adverb 3 rentai#reg#obj
4 obj Ohead) (f4) Penalty = 1
0 (4adverb 3 rentai#reg#obj
4obj Ohead) Function name rentai# here
pass#obj is a passive usage in which the phrase ``bought'' modifies the adjunct with ``book,'' indicating that ``book'' is the object of ``buy,'' and the function name ``renta.''
In i#reg#obj, the phrase ``bought'' modifies ``hon'', meaning respect, and ``hon'' means ``buy''.
It means that it is the object of.

The most efficient question for selecting one of these four structure expression vectors is that the above method considering only the penalty is based on the ambiguity of the first column of the vector, that is, the ``yesterday'', and the fourth column. In other words, the ambiguity of the meaning of ``reru'' can be considered with the same priority. However, since the dialogue processing unit already has a question pattern information table as shown in Figure 13, by referring to this table, compared to the question ``asking the person involved'', the question ``asking the meaning of the word'' is 0.2
It turns out that there is only so much information. So what is the system responsible for (f5) 4 yesterday? I ask this question. After that, you can ask the person what ``reru'' means, or you can stop asking questions if you think the amount of information is sufficiently small. If the question is stopped, this means that the system has determined that it is unnecessary to separate the meaning of ``reru'', taking into account the operator's grammatical knowledge. In that case, the output function structure does not determine the meaning of "reru" as shown in Figure 12.
In other words, it is compressed. Having obtained this functional structure, the system may select the meaning of "passive" by a known default process if necessary.

Effects As is clear from the above explanation, the present invention has the following effects.

(1) Since it does not require knowledge of the target language, it can be applied to, for example, a Japanese-English machine translation system operated by a Japanese person.

(2) The information that the operator should add is necessary and sufficient to eliminate ambiguity, so there is no need to add unnecessary information, and the order of questions is the most efficient.
The burden on the operator can be reduced.

(3) The content of the information that the operator should add is easy to understand, so
The content of the questions is related to dependencies regarding input sentences and the roles of syntactic elements, and it is possible to eliminate interfaces that are difficult for the operator to understand, such as showing analysis structures.

(4) Does not require the operator to have more grammatical knowledge than necessary.

Also, the answers to questions that require a lot of grammatical knowledge may be wrong. Additionally, the ease of each question is calculated, and easy questions are more likely to be asked first. Furthermore, the simpler the answer to the question, the more important the answer can be.

[Brief explanation of the drawing]

FIG. 1 is a block diagram for explaining an embodiment of a natural language sentence analysis device according to the present invention, and FIG. 2 shows a dictionary, syntax rules,
3 is a flowchart of the processing section of the natural language sentence analysis device according to the present invention; FIG. 4 is a diagram showing the contents of one function-related section; FIG. 5 is a diagram showing the component structure; 6th
7 is a diagram showing the functional structure, FIG. 8 is a diagram showing other component structures, FIGS. 9 and 10 are diagrams showing other functional structures, and FIG. 11 is a diagram showing dialogue. FIG. 12 is a diagram showing still another functional structure, and FIG. 13 is a diagram showing a question pattern information amount table. 1...Input section, 2...Output section, 3...Syntax analysis section,
4... Dialogue processing section, 5... Dictionary section, 6... Syntax rule section, 7... Constraint section, 8... Function related section. Patent applicant Rico Co., Ltd. - Figure 1 Figure 4 Figure 2 (a) Dictionary D (b) Syntax rule R Figure 2 (e) Constraint C Figure 3 Figure 5 P He was playing in the park I3τ Fig. 6 Fig. 7 Fig. 8 I bought a book yesterday. I read it. Ward 9. Fig. 10 Fig. 11 Fig. 12

Claims (1)

[Claims] 1. An input section that inputs a natural language sentence, morpheme-divides the input sentence from the input section, creates a plurality of component structures for the morpheme-divided thing, and A syntax analysis unit that creates a functional structure from each constituent structure and applies constraints when creating the functional structure, and an interaction process that performs ambiguity resolution processing through dialogue after the syntax analysis by the syntax analysis unit. and an output unit that outputs an analysis result by the interaction processing unit, and the interaction processing unit extracts one or more pieces of information from each of the plurality of functional structures to distinguish it from other functional structures, and outputs the information. A natural language sentence analysis device characterized by performing one or more of the following in an interactive manner. 2. An input section that inputs a natural language sentence, divides the input sentence from the input section into morphemes, creates multiple component structures for the morpheme division, and creates a structure from each of the component structures. A syntax analysis unit that creates a functional structure and applies constraints when creating the functional structure; and a syntax analysis unit that extracts one or more features that distinguish each of the multiple functional structures from other functional structures; It consists of a dialogue processing unit that asks one or more of the questions and uses the answers to perform ambiguity resolution processing, and an output unit that outputs the analysis results of the dialogue processing unit, and the dialogue processing unit A natural language sentence analysis device characterized by having information regarding certainty. 3. The syntax analysis unit has a calculation means for calculating a value related to incompatibility with each functional structure, and the interaction processing unit is configured to calculate a value related to incompatibility with each functional structure, and the interaction processing unit is configured to calculate a value related to incompatibility with each functional structure. , the natural language sentence analysis device according to claim 1 or 2, wherein the dialogue is performed based on the value related to the incompatibility.