KR101709185B1 - Method and system for building selectional restriction dictionary using sentence pattern of predicate - Google Patents

Abstract

The present invention provides a method and system for constructing an optional constraint dictionary using sentence type information of a verb. A method for constructing a selection constraint dictionary using sentence type information of a verb, comprising the steps of: obtaining sentence type information of a verb; extracting a verb from the obtained sentence type information; and searching for a use sentence including the extracted verb; Analyzing the meaning of the searched example sentence and filtering a usage sentence including a verb with a different meaning from the verb; analyzing a syntax of the filtered usage sentence to filter the irrelevant constituent components; Selecting a constraint object in a usage sentence, extracting an upper path of the selected constrained object using a lexical meaning network, setting a selection constraint path for the verb through comparison of the extracted upper paths, And constructing a selection constraint dictionary for the verb using the set selection constraint path All.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and system for constructing a selection constraint dictionary using syntax information of a verb,

More particularly, the present invention relates to a method and system for constructing a selection constraint dictionary for understanding the structure and meaning of a sentence using information related to a verb phrase, ≪ / RTI >

Natural language processing technology is a technology that mechanically analyzes natural language used by humans and makes it into artificial language form that computer understands.

Natural language processing techniques are divided into morpheme analysis step, syntax analysis step, semantic analysis step, discourse analysis step, word and sentence generation step.

The parsing step is a step of determining the grammatical structure of the sentence by extracting the constituents constituting the sentence based on the morphological analysis result and analyzing the relation between them.

The semantic analysis step identifies the meaning of the whole sentence by classifying the meaning of the words composing the sentence and analyzing the semantic relation between the constituents of the sentence.

Conventional selection constraint dictionaries required for such a parsing step and a semantic analysis step are a method of manually constructing a linguistic expert and a method of semiautomatically constructing using a fixed semantic system and a corpus.

However, the 'Sejong Electronic Dictionary' using the former method has a problem that the accuracy and the time and cost are very high by constructing manually all the seven hierarchical level and the 581 semantic classification systems (FIG. 1 shows 'Sejong Electronic Dictionary' An example of a semantic classification system).

The latter method is semiautomatically constructed so that the work efficiency is high, but there is a problem that it is difficult to determine the degree of meaning that varies from one company to another.

For example, in the case of a sentence such as 'X takes Y' (where 'Tada' means 'ride on a ride or animal'), the following analysis is made automatically using a corpus.

[Gauguin (X: person) became an apprentice sailor in 1865 and took on merchant ship [Y: Transportation]

The subject 'Gauguin' is sufficient for the meaning of [person]

[The man avoided the woman to have a baby]

It is necessary to classify 'woman' as a meaning of 'woman', which is more subdivided than the meaning of [person].

That is, the latter method can not solve the problem that the degree of semantic classification should be set differently for each verb, and when the semantic classification system is changed, there is a problem that all the information constructed needs to be modified in accordance with the changed system.

In order to solve the above-described problems, the present invention provides a method and system for constructing a selection constraint dictionary using the syntax information of a verb that automatically sets an appropriate notion concept for each verb using a Korean lexical semantic network without fixing the semantic classification system The purpose is to provide.

According to an aspect of the present invention, there is provided a method for constructing a selection constraint dictionary using syntax information of a verb, including: obtaining syntax verb information of a verb using an information acquisition unit; Extracting a vernacular from the acquired verb phrase information using the verb extraction unit; Searching for a usage sentence including the extracted verb using the filtering unit, analyzing the meaning of the searched usage sentence, and filtering a usage sentence including a verb having a different meaning from the verb; Selecting a constrained object in the filtered usage sentence using the path extracting unit, and extracting an upper path of the selected constrained object using the lexical semantic network; And setting a selection constraint path for the verb through comparison of the extracted upper paths using the dictionary construction unit and constructing a selection constraint dictionary for the verb using the set selection constraint path . ,
The filtering further includes parsing the filtered usage sentence to filter components that are not related to the spoken word.
In the step of filtering the usage sentence, the retrieval of the usage sentence including the term is performed based on the text big data.
The selection constraint object in the step of extracting the upper path is a constituent element serving as a noun in the filtered and remaining sentence.
The step of constructing the selection constraint dictionary may further include the step of collecting additional usage sentences when the set constraint path has not been denied, and verifying the selection constraint path using the collected usage sentences and the lexical semantic network .
Meanwhile, the selection constraint dictionary construction system using the phrase information of the verb according to the present invention includes an information obtaining unit for obtaining the phrase information of the verb; A vernacular extraction unit for extracting a vernacular form from the acquired sentence type information; A filtering unit for searching for a usage sentence including the extracted pronoun extracted from the spoken word extracting unit and analyzing the meaning of the searched used sentence to filter a usage sentence including a verb with a different meaning from the vernacular; A path extracting unit selecting a selection constraint object in the filtered example sentence and extracting an upper path of the selected constrained object using a lexical semantic network; And a dictionary building unit for setting a selection constraint path for the verb through comparison of the extracted upper paths and constructing a selection constraint dictionary for the verb using the set selection constraint path.
The filtering unit analyzes the syntax of the filtered usage sentence to filter components not related to the spoken language.
The filtering unit searches for a usage sentence including a verb based on the text big data.
The selection constraint object in the path extractor is a constituent element serving as a noun in the filtered and remaining sentence.
If there is a constituent element of which the set constraint path is not set, the dictionary constructor further collects a usage sentence and verifies the selection constraint path using the collected usage sentence and the lexical semantic network.

According to the present invention, the selection constraint dictionary is automatically constructed using the phrase information of the verb phrases, thereby providing an advantage that the time and cost required for the construction can be saved. In addition, it is possible to eliminate the hassle of reconstructing the selection constraint information according to the change of the lexical semantic network.

1 is a diagram illustrating an example of a semantic classification system of the Sejong electronic dictionary according to the prior art.
2 is a flowchart illustrating a method of constructing a selection constraint dictionary using syntax information of a verb according to an embodiment of the present invention.
FIG. 3 is a block diagram illustrating a configuration of a selection constraint dictionary building system using a selection constraint dictionary building method using sentence type information of a verb according to an embodiment of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS The above and other objects, advantages and features of the present invention and methods of achieving them will be apparent from the following detailed description of embodiments thereof taken in conjunction with the accompanying drawings.

The present invention may, however, be embodied in many different forms and should not be construed as being limited to the exemplary embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, And advantages of the present invention are set forth in the appended claims.

It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The singular forms herein include plural forms unless the context clearly dictates otherwise. &Quot; comprises "and / or" comprising ", as used herein, is intended to include the use of one or more other components, steps, operations, and / And does not exclude the presence or addition of a compound.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

2 is a flowchart illustrating a method of constructing a selection constraint dictionary using syntax information of a verb according to an embodiment of the present invention.

As shown in FIG. 2, a selection constraint dictionary construction method using sentence type information of a verb according to an embodiment of the present invention acquires sentence type information of a verb from a corpus (S200). Here, the corpus may be information obtained by various kinds of sentences included in the electronic dictionary.

For example, when constructing a selection constraint dictionary for a verb "Tada" having the meaning of "putting the body on a ride or animal's back," the sentence information for the verb "Tada" (' 'To ride' etc).

When the verb form information of the verb is obtained, a verb is extracted from the sentence form information of the verb (S210).

For example, if the sentence type information of the acquired vernacular is 'to ride on' or 'to ride on', the vernacular 'ride' can be extracted here.

If the vernacular is extracted, the usage sentence including the extracted vernacular is searched (S220), and the meaning of the searched use sentence is analyzed to filter the usage sentence including the verb with a different meaning (S230).

At this time, the meaning of the usage sentence or the usage sentence can be retrieved on the basis of the text big data, and the technique of analyzing the meaning of the sentence is a known technology, so a detailed description will be omitted.

For example, the result of searching for a usage sentence including the extracted vernacular 'Tada' is as follows,

1. [Gauguin] became an apprentice sailor in 1865 and took [Merchant]

2. [Sungjeongjeon] is in [fire] during the period

As a result of analyzing the meanings of the examples 1 and 2, the usage sentence 2 can be filtered because it is a usage sentence including a vernacular "Tada" having a meaning different from "Tada" of "putting on the body of a vehicle or animal" have.

If a usage sentence including a verb with a different meaning is filtered out among the used sentence sentences, the syntax of the filtered usage sentence is analyzed and the constituent components irrelevant to the verb extracted from the usage sentence are filtered (S240).

For example, the analysis of the phrase of the above-mentioned example sentence 1 that has been filtered out shows that if the sentence is judged to be constitutional components that are not related to '1865', 'apprentice sailor', ' Can be filtered.

When the filtering process is completed, a selection restriction object is selected in the filtered usage sentence (S250). Here, the selection constraint object may be a component that acts as a 'noun' in the remaining sentences that are filtered.

For example, if the result of the filtering process is as follows,

3. [Gauguin] got on [Merchant]

4. [They] are in [the car].

5. [she] was in [elevator].

6. [Safety circle] in [Motorcycle]

7. Take the airplane

8. To [the passenger] on [the bus]

The selection constraints to be selected are 'Gogeng', 'Merchant', 'They', 'Cha', 'She', 'Elevator', 'Motorcycle', 'Safety Zone', 'Airplane', 'Bus' ≪ / RTI >

If the selection constraint object is selected, the upper path of the selection constraint target for each investigation is extracted using the lexical meaning network (S260). At this time, the vocabulary may be Korean KorLex or UwordMap.

For the above example, the upper path extracted for the investigation 'a' and 'a' may be as shown in Table 1.

division Top path (→) this them man Person animal Creature she Woman Person animal Creature Safety circle Person animal Creature passenger customer user Person animal Creature on car Transportation Machinery machine elevator Conveying machine Machinery machine Motorcycle bicycle garden truck car Transportation Machinery machine airplane aircraft Transportation Machinery machine Bus Large car car car Transportation Machinery machine Front Seat place space

In the above case, the upper path is extracted only for the search 'i' and 'e', but when a usage sentence including ', a, b, a' is searched for, Of course, be extracted.

When the upper path is extracted, a common path is extracted from the extracted upper paths, and the extracted common paths are compared to set a common path having the longest common path as the final selection constraint path (S270).

For example, in the extracted selection constraint path, the final selection constraint path for the search 'i' may be determined as' person → animal → organism ', and the final selection constraint path for the' '.

In this case, when there is a component whose matched selection constraint path is not matched such as 'elevator' and 'front edge', the step of collecting additional usage sentences and the step of collecting the used sentence and the lexical meaning network, The method comprising the steps of:

When the final selection constraint path is set, the selection constraint dictionary is constructed using the information about the set selection constraint path (S280), and semantic information corresponding to the sentence type of the specific verb can be provided to the user.

FIG. 3 is a block diagram illustrating a configuration of a selection constraint dictionary building system using a selection constraint dictionary building method using sentence type information of a verb according to an embodiment of the present invention.

The selection constraint dictionary construction system 100 using the selection constraint dictionary construction method according to an embodiment of the present invention includes an information acquisition unit 110, a spoken word extraction unit 120, a filtering unit 130, a path extraction unit 140, And a dictionary construction unit 150. [

The information obtaining unit 110 obtains the sentence type information of the verb in the corpus and provides the word extracting unit 120 with data on the sentence type information of the verb. Here, the corpus is information including various types of sentences, and the information obtaining unit 110 can obtain the sentence type information of the verb in the electronic dictionary 10.

The vernacular extraction unit 120 extracts a verb from the acquired vernacular shape information by receiving the data on the verb form verb information from the information obtaining unit 110 and supplies the extracted verb data to the filtering unit 130 do.

The filtering unit 130 receives the data about the verbal words from the word extracting unit 120, searches the usage sentences including the extracted verbs, analyzes the meaning of the searched usage sentences, and includes a verb with a different meaning from the extracted verb Filtered sentences are used.

At this time, the filtering unit 130 can retrieve the meaning of the usage sentence or the usage sentence based on the text big data 20, and the technique of analyzing the meaning of the sentence is a known technology, and a detailed description thereof will be omitted.

When a usage sentence including a verb of another meaning is filtered, the filtering unit 130 analyzes the syntax of the filtered usage sentence, filters constituent elements not related to the verb extracted from the usage sentence, To the path extracting unit (140).

The path extracting unit 140 receives the filtered usage sentence data from the filtering unit 130 and selects a selection constrained object from the filtered usage sentence. Here, the selection constraint object may be a component that acts as a 'noun' in the remaining sentences that are filtered.

When selection of the selection constraint object is completed, the path extracting unit 140 extracts the upper path of the survey-specific selection constraint object using the lexical meaning network 30, And provides it to the building unit 150. At this time, the vocabulary may be Korean KorLex or UwordMap.

The dictionary construction unit 150 receives the data on the upper path of the selected constraint metabolism extracted from the path extraction unit 140, extracts common paths from the extracted upper paths, compares the extracted common paths, The common path having the longest path among the paths is set as the final selection constraint path.

At this time, if there is a component whose matched selection constraint path does not match, the dictionary construction unit 150 further collects usage sentences and verifies the selection constraint path using the collected usage sentences and the lexical semantic network 30 can do.

When the final selection constraint path is set, the dictionary construction unit 150 constructs a selection constraint dictionary using information on the set selection constraint path, and provides semantic information corresponding to the sentence type of the specific verb to the user.

According to the present invention, it is possible to efficiently construct selection constraint information for a spoken word and to provide an advantage that it is not necessary to reconstruct predetermined selection constraint information even if the vocabulary semantic network is changed according to various purposes.

In this way, the natural language question answering system, the text mining technique, and the text mining technique can be realized by automatically expanding or verifying the selection constraint dictionary to understand the text at the semantic level based on the syntactic structure or semantically attached sentence, It can be used as an important technology for Semantic Based Information Retrieval, Machine Translation system, and Information Extraction.

By enhancing the accuracy of the sentence unit analysis technique, the parsing text and the semantic recognition technology, the semantic-based knowledge service is activated, thereby providing an advantage of increasing the utilization of the entire text big data.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, but, on the contrary, is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the invention.

Therefore, the embodiments of the present invention are not intended to limit the scope of the present invention, and the scope of the present invention is not limited by these embodiments. It is intended that the present invention cover the modifications and variations of this invention provided they come within the scope of the appended claims and their equivalents.

Claims (10)

Acquiring sentence type information of a verb using an information acquisition unit;
Extracting a vernacular from the acquired verb phrase information using the verb extraction unit;
Searching for a usage sentence including the extracted verb using the filtering unit, analyzing the meaning of the searched usage sentence, and filtering a usage sentence including a verb having a different meaning from the verb;
Selecting a constrained object in the filtered usage sentence using the path extracting unit, and extracting an upper path of the selected constrained object using the lexical semantic network; And
Setting a selection constraint path for the verb through comparison of the extracted upper paths using a dictionary building unit, and constructing a selection constraint dictionary for the verb using the set selection constrained path
The method comprising the steps of:
The method according to claim 1,
Wherein the filtering further comprises analyzing a phrase of the filtered example sentence to filter out components not related to the verb phrase.
The method according to claim 1,
Wherein the step of filtering the usage sentence includes searching for a usage sentence including a verb based on the text big data.
The method according to claim 1,
Wherein the selection constraint object in the step of extracting the upper path is a constituent element serving as a noun in the filtered and remaining sentence.
The method according to claim 1,
The step of constructing the selection constraint dictionary may further include the step of collecting additional usage sentences when the set constraint path has not been denied, and verifying the selection constraint path using the collected usage sentences and the lexical semantic network The method further comprising the steps of:
An information acquisition unit for acquiring sentence type information of a vernacular;
A vernacular extraction unit for extracting a vernacular form from the acquired sentence type information;
A filtering unit for searching for a usage sentence including the extracted pronoun extracted from the spoken word extracting unit and analyzing the meaning of the searched used sentence to filter a usage sentence including a verb with a different meaning from the vernacular;
A path extracting unit selecting a selection constraint object in the filtered example sentence and extracting an upper path of the selected constrained object using a lexical semantic network; And
A dictionary building unit for setting a selection constraint path for the verb through comparison of the extracted upper paths and constructing a selection constraint dictionary for the verb using the set selection constraint path;
A selection constraint dictionary construction system using the phrase information of a vernacular dictionary.
The method according to claim 6,
Wherein the filtering unit analyzes the syntax of the filtered example sentence and filters the component not related to the spoken word.
The method according to claim 6,
Wherein the filtering unit searches the usable sentence including the verbal word based on the text big data.
The method according to claim 6,
Wherein the selection constraint object in the path extracting unit is a constituent element serving as a noun in the filtered remaining sentence.
The method according to claim 6,
Wherein the dictionary construction unit further collects usage sentences and verifies the selection constraint path using the collected usage sentence and lexical semantic network when the set selection constraint path includes non-denominated components, A selection constraint dictionary construction system using.

Images