JPH02188870A - Context processor - Google Patents

Abstract

PURPOSE:To effectively and accurately decide a context structure even in case the context structure includes the unregistered words by noting the rhetoric expressions emerging in a text to obtain a rhetoric structure and analyzing the context structure of the text based on the rhetoric structure. CONSTITUTION:A corresponding table memory part 5 stores a corresponding table which shows the rhetoric expressions and the connecting relations shown by the rhetoric expressions. A connecting relation extracting part 4 is added together with a rhetoric structure deciding part 6. Then the connecting structure of the connecting relation among the sentence parts is checked from the rhetoric expressions for decision of a rhetroic structure. A context structure is decided based on the rhetoric structure. Thus it is possible to effectively decide a context structure even though the information on the nouns and verbs are insufficient. Furthermore a proper context process is secured since the context structure is checked based on the rhetoric structure secured among sentence parts.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to a context processing device that can effectively determine the context structure of a text in natural language processing.

(Prior Art) In recent years, various researches on natural language processing have been carried out, and efforts have even begun to be made to examine the context structure of text and analyze its meaning. Traditionally, this type of context structure determination has focused on examining the nouns and verbs that appear in the text.
Modifications regarding these words are performed while checking the relationships among the nouns and actions by referring to a dictionary that stores information such as conditions and relationships between parts of speech. In other words, the context structure is determined using the nouns and verbs that appear in the text as clues.

However, it is important to register information about all nouns and verbs that may appear in a text in a dictionary. For this reason, it has become necessary to create information on nouns and verbs used in the target field each time, depending on the content mentioned in the text, for example, and to register this information in a word dictionary. However, the work itself is very difficult, and it has been difficult to adapt it to texts in various target fields and use it as is to determine the context structure.

Additionally, texts often use rhetorical structures such as premises and conclusions, or judgments and illustrations, between multiple paragraphs. However, as mentioned above, conventionally, only the nouns and verbs in the text are used as cues to examine the context structure.
Sometimes the required context structure did not match the above-mentioned rhetorical structure and was inappropriate.

(Problem to be solved by the invention) In this way, in the past, only the nouns and verbs that appear in a text are used as clues to examine the context structure. It is an object of the present invention to provide a highly practical context processing device that can satisfactorily determine the context structure of some texts, such as the need to register information regarding nouns and verbs in a dictionary in advance.

[Structure of the Invention] (Means for Solving the Problems) The context processing device according to the present invention can process conjunctions, etc. that appear frequently in individual texts, even though they occur less frequently than nouns and verbs. Focusing on the rhetorical expressions in the text, we prepare a correspondence table that shows the rhetorical expressions and the connection relationships that the rhetorical expressions mean, and then refer to this correspondence table to determine the relationship between the sentence parts that make up the text based on the rhetorical expressions that appear in the text. find the rhetorical structure of the text by examining the connectivity of
The present invention is characterized in that the context structure of the text is determined using this rhetorical structure.

Furthermore, a relational table indicating connection rules between the connection relations implied by the rhetorical expression is prepared, and a connection structure (structure) of a series of connection relations between sentence parts constituting the text is obtained by referring to the correspondence table. According to the present invention, even if all the information about nouns and verbs that appear in a text is not registered in a dictionary, the connections between sentence parts can be determined from the rhetorical expressions by focusing on the rhetorical expressions in the text. It examines the connection structure of relationships to determine its rhetorical structure, and determines the context structure according to this rhetorical structure, so it is possible to effectively determine the context structure even when information about nouns and verbs is insufficient. becomes. Furthermore, since the context structure is examined by utilizing the rhetorical structure between sentence parts, it becomes possible to perform appropriate context processing.

(Embodiment) Hereinafter, a context processing device according to an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a schematic diagram of the main parts of the apparatus according to the embodiment. Reference numeral 1 denotes a text input section for inputting text such as document data or recorded data to be analyzed. The format structure analysis section 2 executes morphological analysis of sentences containing format information such as chapters and sections contained in the text inputted from the text human resource section 1. Through this morphological analysis, morphemes such as nouns and verbs in the text are individually determined.

The processing functions described above exhibit the same functions as those of conventional context processing devices, and are realized using various techniques that have been proposed in the past.

Here, the feature of this device is that according to the morphological analysis results of the text obtained by the morphological analysis section 3,
The apparatus is provided with a connection relation extraction section 4 that uses rhetorical expressions in the text as clues to find connection relations between a plurality of sentence parts (for example, bar graphs) that make up the text.

For example, as shown in FIG. 2, the connection relation extraction unit 4 refers to a correspondence table storage unit 5 that stores a correspondence table showing rhetorical expressions such as conjunctions and their connection relations, and examines the rhetorical expressions in the text. This is to find the connection relationship between baragraphs etc. that are connected by the rhetorical expression. For example, when the rhetorical expression "however" is required, a connection relationship indicating that the rhetorical expression is in a "supplementary" relationship with the peak 5 that precedes the rhetorical expression is determined.

The connection relation extraction unit 4 sequentially obtains all rhetorical expressions appearing in the text, and also obtains individual connection relations between sentence parts indicated by the rhetorical expressions.

In addition to the connection relation extraction section 4, another feature of this apparatus is that it includes a rhetorical structure determination section B. For example, as shown in FIG. 3, the rhetorical structure determining section B refers to a relational table storage section 7 that stores, as a relational table, prohibition rules relating to the connections indicated by the above-mentioned rhetorical expressions. The rhetorical structure within a baragraph and between baragraphs is determined from the series of connection relationships between sentence parts found in the above.

This relationship table contains information indicating structural connection prohibition rules for the connection relationship shown by rhetorical expressions as shown in Figure 3(a), and information indicating the negative rules for the connection relationship as shown in Figure 3(b). It consists of a table in which information indicating connection rules is registered, and shows a structure between connection relationships that is inappropriate as a rhetorical structure of a text.

The context structure is determined by following the rhetorical structure of the text as described above, or by determining the rhetorical structure of the text, using knowledge of the connections and dependencies of the parts of speech that appear in the text. In other words, this sentence shows an unusual combinatorial structure that is almost always used in -morphisms.

The rhetorical structure determination section B refers to the relational table storage section 7 having such information, and determines the rhetorical structure of the text from the series indicating the connection relations between bar graphs connected by the rhetorical expressions obtained by the connection relation extraction section 4. The rhetorical structure of the text is determined by finding the combination structure while eliminating inappropriate combinations. Note that if the rhetorical structure of the text is not uniquely determined and a plurality of rhetorical structure candidates are found, the decision is left to the next context structure determining section 8.

Therefore, the context structure determination section 8 basically consists of the knowledge storage section 9
According to the knowledge stored in the text, such as information about the field to which the content mentioned in the text belongs, and the context structure of the text found in this way, the application part 1
0 executes predetermined text editing processing such as creating a summary.

The above is the configuration of this device and its functions.

The process of determining a rhetorical structure based on the above-mentioned rhetorical expressions will now be described in more detail with reference to FIGS. 4 and 5.

FIG. 4 is a diagram showing an example of text to be subjected to context processing. When such a text is input, the format structure analysis unit 2 divides the sentence part constituting the text into a plurality of paragraphs or sentences based on format information such as indentations, punctuation marks, commas, periods, etc. Here is paragraph $11. #2. Find each of ~#5. The morphological analysis unit 3 executes morphological analysis for each paragraph obtained in this way.

Based on the morphological analysis results of each paragraph, the connection relation extraction section 4 searches the rhetorical expressions in each paragraph by referring to the correspondence table storage section 5, and for the detected rhetorical expressions, the rhetorical table shown in FIG. As shown in Figure 5A, the text shown in Paragraph #1 has the following connection relationship [1 contrast 2 direct conjunction 3 direct conjunction 4 reference 5]. #2. ~#5 is required to be connected in sequence.

However, such a connection relationship series merely indicates the connection relationships between multiple paragraphs, but does not show how their rhetorical structures are combined.

However, when considering the combination of the above connection relation series that forms a rhetorical structure, it is required that these five paragraph connection relations are "contrastive." Furthermore, from paragraph s3, the rhetorical expression ``these'' and their connection relation ``conjunction'' are required. Similarly, from paragraph #4, the display expression "for this reason" and the connection relationship "conjunct",
From paragraph #5, below r there is the rhetorical expression "that" and "
A connection relationship called "reference" is required.

Based on such a connection relationship, the connection relationship extraction unit 4 can have two types of combination structures. and 4
In the case of one paragraph, there are five ways, and the number of forms that the combination structure can take is theoretically determined by the number of paragraphs.

The combinatorial structure of this series of connections is where each sentence part (paragraph) in the text is located.

For example, in the rhetorical structure (combinative structure) [1 contrast [[2 injunction 3] injunction [4 reference 5]]], the first paragraph #1 is the first It shows that there is a "contrastive" relationship with the entirety of paragraph #2 of paragraph #2 to paragraph #5 of fifth paragraph. Regarding the second paragraph #2 to the fifth paragraph #5, the third paragraph #3 is derived from the second paragraph #2 in a "conjunctive" relationship, and based on these, the fourth paragraph It is shown that paragraph #4 is derived in a "conjunctive" relationship, and that the fifth paragraph #5 is attached to the fourth paragraph #4 in a "reference" relationship, and this fourth paragraph #5 is attached to the fourth paragraph #4 in a "reference" relationship. This indicates that a portion related to the content of paragraph II4 exists in another portion of the text.

This expression of rhetorical structure indicates the connection structure relationship between multiple paragraphs. However, the combinational structure adopted by the series of connection relationships shown in B above simply represents the structural form of the combination, and some of them are inappropriate as rhetorical structures in the sentence structure. Therefore, the rhetorical structure determination section 6 refers to the relational table storage section 7 to find a structure in an inappropriate rhetorical relationship among the combination structures, and excludes this structure from the rhetorical structure candidates.

For example, as shown in FIG. 5C, the connection prohibition rule "Order connection [... contains is detected to be inappropriate as a rhetorical structure.

Therefore, the combinatorial structure of the connection relation series including the combination structure of "order tangent [3 order tangent"] [1 contrast [[2 order tangent [3 order tangent [4 order tangent]
Reference ] ] 5] 1 is excluded from the candidates for its rhetorical structure.

Similarly, the negative connection rule "contrast [・
...Exclude the combinatorial structure of the connection relation series that includes the combination ``contrast [[2 order connection J'' from [X order connection] from the rhetorical structure candidates. In some cases, the series of combinations may be left as rhetorical structure candidates.

In this way, the rhetorical structure determining section 6 selects inappropriate rhetorical structures from among the combinations of rhetorical structures that the series can theoretically take, according to the series of connective relationships determined by the connective relationship extracting section 4. are detected and excluded by referring to the relational table storage unit 7, and the candidates as rhetorical structures are narrowed down. If a rhetorical structure is uniquely found in this narrowing down of candidates, this is determined as the rhetorical structure, and if multiple candidates remain, the decision is made by determining the context structure using knowledge. Leave it to Department 8.

Then, taking into consideration the rhetorical structure of the text determined in this way, the structure of the text content, etc. is determined, and the final context structure is determined.

According to this device configured in this way, it not only analyzes nouns and verbs in a text, but also analyzes the rhetorical structure of the text indicated by the rhetorical expressions that appear in the text, and takes this rhetorical structure into consideration. Since the above-mentioned context structure of the text is determined and context logic is performed, even if information about nouns and verbs that appear in the text is not registered in the dictionary, appropriate information can be determined without being influenced by the unregistered words. It becomes possible to perform context processing. Furthermore, since context processing is performed using rhetorical structures analyzed from rhetorical expressions, it is possible to accurately capture and context-process rhetorical structures such as premises and conclusions, judgments and illustrations, and the context can be appropriately processed. It becomes possible to execute language processing that is accurately captured.

It should be noted that the present invention is not limited to the above-described embodiments, and format structure analysis, morphological analysis, etc. may be performed using various conventionally proposed algorithms as appropriate. Furthermore, the information registered in the correspondence table and relational table is not limited to the above-mentioned example. Furthermore, although Japanese processing has been described here, context processing for other languages such as English can be similarly applied. In addition, the present invention can be implemented with various modifications without departing from the gist thereof.

[Effects of the Invention] As explained above, according to the present invention, the rhetorical structure of the text is determined by focusing on the rhetorical expressions that appear in the text, and the context structure of the text is analyzed using this rhetorical structure. Even in cases where unregistered words are included in a text, the context structure can be determined effectively and accurately, which provides a great practical effect.

[Brief explanation of the drawing]

The figures show a context processing device according to an embodiment of the present invention. FIG. 1 is a schematic diagram of the main parts of the embodiment device, FIG. 2 is a diagram showing an example of the configuration of a correspondence table, and FIG. 3 is a diagram showing the relationship between FIG. 4 is a diagram showing an example of the structure of a table, FIG. 4 is a diagram showing an example of a text subjected to context processing, and FIG. 5 is a diagram schematically showing a form of text rhetorical structure analysis processing. ■...Text input section, 2...Format structure analysis section, 3
...Morphological analysis unit, 4...Connection relationship extraction unit, 5...
・Correspondence table storage unit, B...rhetorical structure determination unit, 7.
...Relationship table storage section, 8...Context structure determination section, 9
...knowledge storage section, IO...application section. Applicant Director of the Agency of Industrial Science and Technology Kozo Iizuka Figure 1 (a) Figure 1

Claims (3)

[Claims] (1) A correspondence table showing rhetorical expressions and the connection relationships implied by the rhetorical expressions is provided, and the connection relationships between the sentence parts that make up the text are checked from the rhetorical expressions appearing in the text by referring to this correspondence table. A context processing device comprising: means for determining the rhetorical structure of the text; and means for determining the context structure of the text using the rhetorical structure. (2) A correspondence table showing rhetorical expressions and connection relations implied by the rhetorical expressions, and a relation table showing connection rules between connection relations, and by referring to the correspondence table, the rhetorical expressions appearing in the text are means for determining a series of connection relationships between sentence parts constituting the text; means for determining the rhetorical structure of the text by referring to the relationship table and examining the connection/structure of the series of connection relationships; and the rhetorical structure of the text. 1. A context processing device comprising: means for determining a context structure of the text using the method. (3) The context processing device according to claim (2), wherein the relationship table is made up of connection prohibition rules between connection relationships, and is used to eliminate connection structures in which the series of connection relationships cannot be taken.