JP2017220009A - Analysis program, analysis method, and analyzer - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an analysis program capable of efficiently identifying a previous sentence required for analyzing a sentence.SOLUTION: The analysis program causes computer to execute a series of processing including: determining, with respect to plural sentences as analysis objects, whether a sentence of analysis object includes a specific expression; and when determining a second sentence which is determined to include the specific expression is included in sentences preceding a first sentence as the analysis object, and when the meaning of a predicative of the first sentence matches with the meaning of a predicative of the second sentence, identifying the second sentence as a reference object for analyzing the first sentence.SELECTED DRAWING: Figure 3

Description

  The present invention relates to an analysis program, an analysis method, and an analysis apparatus.

  Systems that execute natural language processing using object-oriented ideas have been studied (for example, Non-Patent Document 1). Basically, nouns are registered as objects. Generally, adjectives correspond to object attributes. The attribute of the object changes depending on the action or event represented by the verb.

  For example, for the sentence “Taro handed a book to Hanako”, Taro, Hanako, and the book are registered as objects. The book falls from Taro's property list attribute, and the book enters Hanako's property list attribute. Also, the book owner attribute changes from Taro to Hanako.

  Such processing is executed in the order in which the sentences appear. However, certain kinds of sentences cannot be immediately processed as attributes and need to be referred to in trigger processing. Here, the trigger process refers to a process related to consistency check with an already-existing sentence, evaluation of various analyzes, and the like, which are executed for the sentence in response to the appearance of the sentence.

  For example, a sentence including an auxiliary particle such as “only” or “only” corresponds to a sentence that needs to be implemented as a trigger process, such as “Taro can only swim 50 meters in a class”. That is, when a sentence “Jiro swam 100m” later appears, it is necessary to check “Is Jiro the same class as Taro?” (When you see Taro's sentence, you cannot do anything.).

  Also, in the predicate analysis of the sentence "Jiro is ... Saburo ... 100m swim", if Jiro or Saburo is the subject of "Swim", Jiro is in the same class as Taro and Saburo is not , You can use information that Jiro should not be the subject.

  Similarly, a sentence including the highest level expression such as “Taro is the tallest in the class” needs to be implemented as a trigger process.

Japanese Patent Laid-Open No. 07-175808 Japanese Patent Laid-Open No. 04-220767 JP-A 63-305464 JP 2011-013776 A

Hideki Ito. Dependency / word sense ambiguity resolution / anaphora analysis / predicate analysis system based on object-oriented approach. Research report Natural Language Processing (NL), 2014-NL-218 (1), 1-7 (2014)

  However, if the trigger processing is simply implemented, every time each sentence appears, all the sentences that have appeared before are checked, which increases the calculation time.

  Therefore, in one aspect, an object of the present invention is to improve the efficiency of specifying an existing sentence necessary for sentence analysis.

  In one aspect, the analysis program sets a plurality of sentences to be analyzed one by one, determines whether or not the sentence to be analyzed includes a specific expression, and determines whether the sentence before the first sentence to be analyzed If there is a second sentence determined to contain the specific expression, if the meaning of the predicate of the first sentence matches the meaning of the predicate of the second sentence, the second sentence A computer is caused to execute a process for specifying a sentence as a reference object for the analysis of the first sentence.

  As one aspect, it is possible to increase the efficiency of specifying the already-existing sentence necessary for sentence analysis.

It is a figure which shows the hardware structural example of the analyzer in embodiment of this invention. It is a figure which shows the function structural example of the analyzer in embodiment of this invention. It is a flowchart for demonstrating an example of the process sequence which an analyzer performs.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a diagram illustrating a hardware configuration example of an analysis apparatus according to an embodiment of the present invention. The analysis device 10 in FIG. 1 includes a drive device 100, an auxiliary storage device 102, a memory device 103, a CPU 104, an interface device 105, and the like that are mutually connected by a bus B.

  A program that realizes processing in the analysis apparatus 10 is provided by the recording medium 101. When the recording medium 101 on which the program is recorded is set in the drive device 100, the program is installed from the recording medium 101 to the auxiliary storage device 102 via the drive device 100. However, the program need not be installed from the recording medium 101 and may be downloaded from another computer via a network. The auxiliary storage device 102 stores the installed program and also stores necessary files and data.

  The memory device 103 reads the program from the auxiliary storage device 102 and stores it when there is an instruction to start the program. The CPU 104 executes a function related to the analysis device 10 according to a program stored in the memory device 103. The interface device 105 is used as an interface for connecting to a network.

  An example of the recording medium 101 is a portable recording medium such as a CD-ROM, a DVD disk, or a USB memory. An example of the auxiliary storage device 102 is an HDD (Hard Disk Drive) or a flash memory. Both the recording medium 101 and the auxiliary storage device 102 correspond to computer-readable recording media.

  FIG. 2 is a diagram illustrating a functional configuration example of the analysis apparatus according to the embodiment of the present invention. 2, the analysis apparatus 10 includes a sentence analysis unit 11, a basic synonym original form acquisition unit 12, a trigger target specifying unit 13, a trigger processing unit 14, a trigger determination unit 15, a sentence registration unit 16, and the like. Each of these units is realized by processing that one or more programs installed in the analysis apparatus 10 cause the CPU 104 to execute. The analysis apparatus 10 also uses the basic synonym original form dictionary storage unit 17, the sentence arrangement storage unit 18, and the like. Each of these storage units can be realized using, for example, a storage device that can be connected to the auxiliary storage device 102, the memory device 103, or the analysis device 10 via a network.

  The sentence analysis unit 11 executes existing natural language processing (for example, predicate term analysis and anaphora analysis) for each sentence belonging to the sentence set. As a result of the processing by the sentence analysis unit 11, for example, data indicating the semantic structure of the sentence (hereinafter referred to as “semantic data structure”) is output. A sentence set is a set (for example, sentences) of a plurality of sentences having a semantic context. The sentence is plain text in Japanese. Each sentence is subject to analysis one sentence at a time in the order of context in the sentence set.

  The predicate term analysis is detailed in “Hajic et al. The CoNLL-2009 Shared Task: Syntactic and Semantic Dependencies in Multiple Languages. In Proceedings of the Thirteenth Conference on Computational Natural Language Learning, 2009”.

  The basic synonym original form acquisition unit 12 acquires the basic synonym original form of the predicate of each sentence from the basic synonym original form dictionary storage unit 17. In this embodiment, the basic synonym original form is an original form of a word that is typically (generally) used as an expression of a concept that a predicate (verb, adjective, etc.) means. For example, the basic synonym of “jogging”, “kake”, “run” is “run”. The basic synonym original is data in which each basic synonym original is associated with an expression having the same meaning as the basic synonym original. That is, it can be said that the basic synonym original dictionary is information indicating a result of classifying expressions of a plurality of predicates based on commonality of meanings.

  The trigger-required determination unit 15 determines whether each sentence is a reference target of trigger processing (hereinafter referred to as “trigger-required sentence”). The trigger processing refers to processing for checking consistency with the previous sentence and evaluating various types of analysis, which is executed for the sentence according to the appearance of the sentence. In the present embodiment, it is determined that a sentence including a specific expression is a trigger sentence. The sentence including a specific expression is, for example, a sentence including an auxiliary particle such as “only” or “shika” or a sentence including a superlative expression. That is, it is determined that a sentence including an expression meaning limitation is a trigger required sentence.

  The trigger target specifying unit 13 specifies, for each sentence, a sentence before being referred to in the trigger process (trigger required sentence).

  The trigger processing unit 14 refers to the trigger required sentence for the sentence for which the trigger required sentence is specified by the trigger target specifying unit 13 and performs trigger processing on the sentence. In the present embodiment, the contents of the trigger process are not limited to specific ones. The contents of the trigger processing may differ depending on what kind of analysis the analysis device 10 is intended for for each sentence.

  The sentence registration unit 16 stores a record including a semantic data structure, a trigger check flag, and a basic synonym original form in the sentence array storage unit 18 for each sentence. The trigger check flag is information indicating a determination result by the trigger required determination unit 15 and can take a value of 0 or 1. 0 indicates that the trigger statement is not required. 1 indicates a trigger required sentence. Note that the basic synonym original form may be stored in the sentence array storage unit 18 only for the trigger required sentence.

  Each sentence may be analyzed as disclosed in Non-Patent Document 1. That is, the noun of each sentence may be an object, the adjective may be an attribute of the object, and the attribute of the object may change depending on the verb.

  Hereinafter, a processing procedure executed by the analysis apparatus 10 will be described. FIG. 3 is a flowchart for explaining an example of a processing procedure executed by the analysis apparatus.

  In step S101, the sentence analysis unit 11 reads the first sentence in the unprocessed sentence from the sentence set as an analysis target. The sentence to be analyzed is hereinafter referred to as “target sentence”.

  Subsequently, the sentence analysis unit 11 performs predetermined natural language processing on the target sentence (S102). As a result, the semantic data structure of the target sentence is obtained. In the present embodiment, the semantic data structure is not limited to a predetermined data structure.

  Subsequently, the basic synonym original form extraction unit acquires the basic synonym original form corresponding to the predicate of the target sentence from the basic synonym original form dictionary storage unit 17 (S103).

  Subsequently, the trigger target specifying unit 13 determines whether or not a record whose trigger check flag is 1 is stored in the sentence array storage unit 18 (S104). That is, it is determined whether or not a trigger required sentence is detected before the target sentence. If the target sentence is the first sentence, the sentence array storage unit 18 is empty. Therefore, this determination is negative.

  When the corresponding record is not stored in the sentence array storage unit 18 (No in S104), the trigger required determination unit 15 determines whether the target sentence is a trigger required sentence (S105). That is, it is determined whether the target sentence includes the specific expression described above. When the target sentence is a trigger sentence required (Yes in S105), the trigger required determination unit 15 sets 1 to the trigger check flag for the target sentence (S106). On the other hand, if the target sentence is not a trigger required sentence (No in S105), the trigger required determination unit 15 sets 0 to the trigger check flag for the target sentence (S107).

  Subsequent to step S106 or S107, the sentence registration unit 16 registers a record including the semantic data structure, trigger check flag, and basic synonym original form of the target sentence in the sentence array storage unit 18 (S108). That is, the trigger check flag and the basic synonym original form are stored in association with the target sentence. Note that the target sentence itself may also be stored in the record.

  On the other hand, in step S104, when a record whose trigger check flag is 1 (hereinafter referred to as “required trigger record”) is stored in the sentence array storage unit 18 (Yes in S104), the trigger target specifying unit 13 The basic synonym original form of the target sentence is compared with the basic synonym original form of the trigger record required (S109). When the compared two basic synonym original forms match, that is, when the meaning of the predicate of the target sentence matches the meaning of the predicate of the trigger sentence required (S110), the trigger target specifying part 13 Is identified as a statement to be referenced in the trigger process. Therefore, the trigger processing unit 14 refers to the trigger required sentence and executes trigger processing for the target sentence (S111).

  If there are a plurality of trigger records that require the target sentence and the basic synonym original form to match, the trigger processing may be executed with reference to the plurality of trigger statements.

Next, the processing procedure of FIG. 3 will be specifically described using the following example sentences A to C.
A “Taro is the only one who can swim 1 km in one set.”
B "Jiro is one pair."
C "He did a 3km long swim."
In this case, the following (1) to (11) are executed.
(1) The sentence A is set as an analysis target (S101), and its semantic data structure is obtained (S102).
(2) “Swim” is acquired as a basic synonym original form of the predicate “Swim” in the sentence A (S103).
(3) Since the sentence of A includes the auxiliary particle “only”, it is determined that the sentence of A is a trigger-required sentence, and the trigger check flag is set to 1 (S105, S106).
(4) A record of {<Meaning data structure of A>, check flag = 1, basic synonym original form = “swim”} is registered in the sentence array storage unit 18.
(5) Steps S101 to S103 are similarly executed for the sentence B. Note that, for a predicate of noun (or pronoun) + auxiliary verb, as in “one set”, the noun (or pronoun) may be acquired as a basic synonym original form as it is.
(6) The sentence array storage unit 18 has a record relating to the sentence of A whose check flag is 1, but the basic synonym original form of the predicate of the sentence of B is the basic synonym original form of the predicate of the sentence of A Since they do not match, trigger processing is not executed.
(7) Since the sentence B is not a trigger required sentence, 0 is set in the trigger check flag (S105, S106).
(8) A record of {<Semantic data structure of B>, check flag = 0, basic synonym original form = “−”} is registered in the sentence array storage unit 18. “-” Indicates that the value of the basic synonym original form is empty. This is because the sentence of B is not a triggering sentence, so that the basic synonym original form does not need to be registered. At this time, an object corresponding to Jiro may be generated, and an attribute indicating “affiliation: 1 set” may be given to the object.
(9) Subsequently, the C sentence is set as a processing target, and natural language processing is performed (S102). For example, an anaphoric analysis is performed for “He” in the sentence of C. There are two interpretations of “hi”: “Taro” in the sentence A and “Jiro” in the sentence B. In the anaphora analysis, the likelihood is evaluated for each interpretation and an evaluation value is given. Here, it is assumed that the same evaluation value is given to each interpretation.
(10) “Swim” is acquired as a basic synonym original form of the predicate “I went far away” in the sentence of S (S103).
(11) Since the basic synonym original form of the sentence C matches the basic synonym original form of the sentence A, the sentence A is referred to and the trigger process (S111) is performed on the sentence C.

  In the trigger process, for example, a consistency check with A may be performed on the result of the anaphora analysis of the sentence C. Then, since the interpretation of “He = Jiro” is inconsistent with the sentence of A, the evaluation value for the interpretation is deducted. As a result, the possibility that the interpretation of “he = Taro” will be selected eventually increases.

  Regarding the above example, in (11), since the reference object of the trigger process can be limited to A (that is, B can be skipped), the amount of calculation is reduced. That is, efficiency is achieved.

  The contents of the trigger process are not limited to the consistency check for the result of anaphora analysis. Depending on the content of the sentence and the purpose of the analysis, the content of the trigger processing varies.

In the above example sentence, an example in which the predicate is a verb has been described, but the present embodiment is also effective when the predicate is an adjective as follows.
A “Suzuki is the only person in the patent department that is strong against personal computers.”
B "Sato is a patent member."
C “He is strong on personal computers.”
In this case, the A sentence is referred to, and the trigger process of the C sentence is executed.

  As described above, according to the present embodiment, it is possible to reduce the calculation amount and calculation time of trigger processing in the interpretation of the meaning of a sentence. That is, it is possible to increase the efficiency of specifying the already-described sentence necessary for the interpretation of the sentence.

  In the present embodiment, the trigger required determination unit 15 is an example of a determination unit. The trigger target specifying unit 13 and the basic synonym original form acquiring unit 12 are examples of the specifying unit. The trigger check flag is an example of information indicating that a specific expression is included.

  As mentioned above, although the Example of this invention was explained in full detail, this invention is not limited to such specific embodiment, In the range of the summary of this invention described in the claim, various deformation | transformation・ Change is possible.

Regarding the above description, the following items are further disclosed.
(Appendix 1)
Analyzing multiple sentences one sentence at a time
Determine whether the sentence being analyzed contains a specific expression,
When there is a second sentence determined to include the specific expression in a sentence before the first sentence to be analyzed, the meaning of the predicate of the first sentence and the second sentence If the meaning of the predicate matches, the second sentence is identified as a reference object for the analysis of the first sentence.
An analysis program characterized by causing a computer to execute processing.
(Appendix 2)
The specific expression is an expression meaning limitation.
The analysis program according to supplementary note 1, characterized by:
(Appendix 3)
Comparing the meaning of the predicate of the first sentence with the meaning of the predicate of the second sentence with reference to information indicating the result of classifying the expressions of the plurality of predicates based on the commonality of meaning;
The analysis program according to appendix 1 or 2, which causes a computer to execute processing.
(Appendix 4)
Analyzing multiple sentences one sentence at a time
Determine whether the sentence being analyzed contains a specific expression,
When there is a second sentence determined to include the specific expression in a sentence before the first sentence to be analyzed, the meaning of the predicate of the first sentence and the second sentence If the meaning of the predicate matches, the second sentence is identified as a reference object for the analysis of the first sentence.
An analysis method characterized by causing a computer to execute processing.
(Appendix 5)
The specific expression is an expression meaning limitation.
The analysis method according to supplementary note 4, characterized by:
(Appendix 6)
Comparing the meaning of the predicate of the first sentence with the meaning of the predicate of the second sentence with reference to information indicating the result of classifying the expressions of the plurality of predicates based on the commonality of meaning;
6. The analysis method according to appendix 4 or 5, which causes a computer to execute the process.
(Appendix 7)
An analysis device for analyzing a plurality of sentences one by one,
A determination unit that determines whether or not the sentence to be analyzed includes a specific expression;
When there is a second sentence determined to include the specific expression in a sentence before the first sentence to be analyzed, the meaning of the predicate of the first sentence and the second sentence If the meaning of the predicate matches, a specifying unit that specifies the second sentence as a reference object for the analysis of the first sentence;
The analysis apparatus characterized by having.
(Appendix 8)
The specific expression is an expression meaning limitation.
The analysis device according to appendix 7, characterized by:
(Appendix 9)
The specifying unit refers to information indicating a result obtained by classifying expressions of a plurality of predicates on the basis of commonality of meanings, and means a predicate of the first sentence and a predicate of the second sentence. Compare with
The analyzer according to appendix 7 or 8, characterized in that.

DESCRIPTION OF SYMBOLS 10 Analysis apparatus 11 Sentence analysis part 12 Basic synonym original form acquisition part 13 Trigger object specific | specification part 14 Trigger processing part 15 Trigger determination part 16 Sentence registration part 17 Basic synonym original form dictionary storage part 18 Sentence arrangement | sequence storage part 100 Drive apparatus 101 Recording Medium 102 Auxiliary storage device 103 Memory device 104 CPU
105 Interface device B bus

Claims (5)

Analyzing multiple sentences one sentence at a time
Determine whether the sentence being analyzed contains a specific expression,
When there is a second sentence determined to include the specific expression in a sentence before the first sentence to be analyzed, the meaning of the predicate of the first sentence and the second sentence If the meaning of the predicate matches, the second sentence is identified as a reference object for the analysis of the first sentence.
An analysis program characterized by causing a computer to execute processing. The specific expression is an expression meaning limitation.
The analysis program according to claim 1. Comparing the meaning of the predicate of the first sentence with the meaning of the predicate of the second sentence with reference to information indicating the result of classifying the expressions of the plurality of predicates based on the commonality of meaning;
The analysis program according to claim 1 or 2, which causes a computer to execute processing. Analyzing multiple sentences one sentence at a time
Determine whether the sentence being analyzed contains a specific expression,
When there is a second sentence determined to include the specific expression in a sentence before the first sentence to be analyzed, the meaning of the predicate of the first sentence and the second sentence If the meaning of the predicate matches, the second sentence is identified as a reference object for the analysis of the first sentence.
An analysis method characterized by causing a computer to execute processing. An analysis device for analyzing a plurality of sentences one by one,
A determination unit that determines whether or not the sentence to be analyzed includes a specific expression;
When there is a second sentence determined to include the specific expression in a sentence before the first sentence to be analyzed, the meaning of the predicate of the first sentence and the second sentence If the meaning of the predicate matches, a specifying unit that specifies the second sentence as a reference object for the analysis of the first sentence;
The analysis apparatus characterized by having.

Images