KR101450795B1 - Apparatus and method for anaphora resolution - Google Patents

Abstract

The present invention relates to a resource description framework (RDF) triple-based anaphora resolution apparatus which comprises: an RDF triple conversion part which converts each sentence included in input text into RDP triples; and an anaphora processing part which finds anaphora triples including an anaphor from a plulrality of triples converted in the RDF triple conversion part, and specifies an antecedent of the anaphor by comparing the anaphora triples with the antecedent triples of the anaphora triples.

Description

[0001] APPARATUS AND METHOD FOR ANAPHORA RESOLUTION [0002]

The present invention relates to an apparatus and method for restoring large terms.

Recently, a technology for translating the meaning of knowledge and information into RDF (Resource Description Framework) triple, which is a computer understandable format, has been studied. The RDF Triple is an international standard administered by the World Wide Web Consortium (W3C) that provides knowledge and information to subjects [resource], predicate [property], object [ ].

Anaphor can not have its own direction, but is specified by another expression in the sentence, that is, the preceding word. Vocabulary is an indispensable element in natural language and forms the basis of language economics. In order for people to accurately identify the antecedents that the term refers to, they need a lot of reading and training, and they also need extra-linguistic knowledge in addition to their linguistic knowledge. Therefore, it is difficult for the computer to find the leading term that the term refers to.

Until now, natural language automatic processing devices have been developed to deal with vocabularies based on the degree of correspondence between the vocabulary and the grammatical qualities such as sex, number, and personality of the leading words, and the meaning qualities of people, objects, places, We specified the leading word. However, it is difficult to specify the leading term corresponding to a large term when there are several candidate leading words having similar grammatical qualities or semantic qualities, Therefore, there is a need for a method of accurately identifying the leading term corresponding to a large term using the context information represented by the RDF triple.

Korea Patent No. 10-1092355 Method for restoring the term (Dec 09, 2011 notice)

SUMMARY OF THE INVENTION It is an object of the present invention to provide an apparatus and method for restoring a context-based vocabulary.

A resource description framework (RDF) triple-based large term restoration apparatus according to an embodiment of the present invention includes an RDF triple transformation unit for transforming each sentence included in an input statement into an RDF triple, And a vocabulary processing unit for searching for a vocabulary triple including a vocabulary among a plurality of triples converted in the vocabulary unit and comparing the vocabulary triple and the preceding triples of the vocabulary triple to identify a preceding word of the vocabulary.

The to-be-term processing unit may extract a triple having a term in a place of an object and an object in a triple term among the plurality of triples.

The large term processing unit can determine the vocabulary including the directive, the normal noun, the pronoun, and the Young's term display information indicating that the subject or object of the RDF triple is empty.

The to-term processing unit may find an associated preceding triple having the same vocabulary or the same qualities as the corresponding vocabulary triple among the preceding triples and specify the leading term of the vocabulary in the related preceding triple.

The to-term processing unit may search for a subject or object including the noun among preceding triples and specify a vocabulary described in the found subject or object as a preceding word when the corresponding term is a form including an instruction and a noun.

If the subject or the object including the noun is not found among the preceding triples, the term terminology processing unit can find a subject or an object that includes a vocabulary having the same qualities as the noun.

In the case where the large term is located at the subject position, the large term processing unit may compare the object of the preceding triples and the large term triple with the predicated kit to find the related preceding triple and specify the subject of the related preceding triple as a leading character have.

When the large term is located at the object place, the large term processing unit compares the predecessor triple with the subject of the large term triple and the predicate kit to find the related preceding triple and specify the object of the related preceding triple as a leading character have.

The vocabulary processing unit may restore the vocabulary triple by replacing the vocabulary with a leading word and store the restored vocabulary triple.

Wherein the large term processing unit comprises a large term extraction unit for extracting a large term term triple including a large term from a plurality of triples converted by the RDF triple conversion unit, And a RDF triple restoring unit for restoring the large term triple by replacing the large term with a leading character.

A method for restoring a large term based on a Resource Description Framework (RDF) triple based on a plurality of RDF triples corresponding to each sentence of an input query Searching for a corresponding term triple having the same vocabulary or qualities as the corresponding term triple among the preceding triples of the same term triple among the inputted plurality of triples, And identifying the preceding term of the vocabulary in the preceding triple.

The step of searching for the large term triple may include the step of, when at least one of a vocabulary including a directive, a common noun, a pronoun, And the Young's Daughter term display information may be information indicating that the subject or object of the corresponding triple is empty.

Wherein the step of finding the related preceding triple is to search for a subject or object that includes the noun in the preceding triples if the corresponding term is a form including an directive and a noun, Vocabulary can be extracted as a leading word.

In the step of finding the related preceding triple, if the subject or object including the noun is not found among the preceding triples, a subject or object including a vocabulary having the same qualities as the noun can be found.

In the step of searching for the related preceding triple, when the corresponding term is located at the subject position, the object of the preceding triple and the corresponding term triple is compared with the predicated kit to find an associated preceding triple having the same vocabulary or qualities as the corresponding term triple .

The step of specifying the preceding word may specify a subject of the related preceding triple as a leading word.

The finding of the related preceding triple may include comparing the subject of the preceding triple with the subject of the corresponding term triple and the predicated kit to find an associated preceding triple having the same vocabulary or qualities as the corresponding term triple .

The step of identifying the preceding word may specify an object of the related preceding triple as a leading word.

The method of restoring a large term may further include restoring the large term triple by replacing the large term with a leading term, and storing the restored large term triple.

According to the embodiment of the present invention, the meaning of the sentence and the context information are represented by the RDF triple, so that the leading word can be accurately specified as compared with the method of determining the leading word based on the degree of matching of the individual grammatical qualities and the meaning qualities.

According to the embodiment of the present invention, it is possible to precisely specify a leading term corresponding to a large term even if the candidate of the leading term having similar grammatical qualities such as sex, number, and personality and the semantic qualities of people, objects, places, have.

Also, according to the embodiment of the present invention, it is possible to precisely specify the leading word corresponding to the Young-Dang term. Therefore, according to the embodiment of the present invention, a better performance can be expected for automatic language processing apparatuses.

1 is a block diagram of a large term recovery apparatus according to an embodiment of the present invention.
2 is a block diagram of a vocabulary processing unit according to an embodiment of the present invention.
3 is a flowchart of a method for restoring a large term according to an embodiment of the present invention.
4 is a flowchart of a method for restoring a large term according to another embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily carry out the present invention. The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In order to clearly illustrate the present invention, parts not related to the description are omitted, and similar parts are denoted by like reference characters throughout the specification.

Throughout the specification, when an element is referred to as "comprising ", it means that it can include other elements as well, without excluding other elements unless specifically stated otherwise.

Hereinafter, an apparatus and method for restoring large terms according to an embodiment of the present invention will be described with reference to the drawings.

1 is a block diagram of a large term recovery apparatus according to an embodiment of the present invention.

Referring to FIG. 1, a large term restoration apparatus 100 converts a plurality of texts constituting text into a Resource Description Framework (RDF) triple. The term restoring apparatus 100 extracts an anchor included in the triple and identifies a leading term corresponding to a larger term in the preceding triple.

The RDF Triple is an international standard administered by the World Wide Web Consortium (W3C) that provides knowledge and information to subjects [resource], predicate [property], object [ ].

A vocabulary can not have its own direction, but is specified by another expression in the sentence, that is, a preceding word. For example, directives, directives + nouns, pronouns, common nouns, and the like can be major terms. Or if a variable such as "? X" or "? Y" appears in place of the subject of the transformed RDF triple and the object, then the empty subject and object may be a zero term term.

The term terminology recovery apparatus 100 includes an input analysis unit 200, an RDF triple conversion unit 300, a term processing unit 400, and an RDF triple repository 500.

The input analysis unit 200 includes an input unit 210, a morphological analysis unit 230, a word generator 250, and a sentence component analysis unit 270.

The input unit 210 receives input text composed of text. The input statement may be composed of a plurality of sentences. The input unit 210 can receive input text composed of text such as a text file and a web document.

The morpheme analysis unit 230 analyzes the input sentence as a morpheme. The morpheme analysis unit 230 analyzes the input sentence using a morpheme analyzer. A morpheme is the smallest unit of meaning that makes up a sentence.

The phrase generator 250 generates a word phrase based on the morpheme. A phrase is a sentence component that is separated by a space in a spelling sentence. The vernacular is divided into cognition (NN), vernacular (VV), affixed verb (VNP), customary (MM), adverb (MA), exclamation (IC), and connective (CONJ)

The sentence component analysis unit 270 analyzes the role in the sentence of the phrase, that is, the sentence component. Sentence components are divided into subject (SBJ), object (OBJ), predicate (PRD), bore (CMP), modifier (MOD), additional word (AJT), connection word (CNJ) and independent word (INT).

The RDF triple conversion unit 300 generates sentence segment information by dividing each sentence into sentence blocks based on the sentence components output from the input sentence analysis unit 200. [ The sentence block is composed of a chan- nel block N, a compound noun block N, a proper noun block P, a unit noun block U, a grammatical block G, a par- ticular connection block O, a verb block V, Block C, adverbial block B, clause block S, and question block Q. [

The RDF triple conversion unit 300 converts each sentence into an RDF triple based on the sentence component and sentence segment information output from the input sentence analysis unit 200. An RDF triple consists of a subject, a predicate kit, and an object.

If there is a term in the RDF triple, the term terminology processing unit 400 specifies a leading term corresponding to the term in order to restore its meaning.

The term term extraction unit 410 searches for a large term triple including a large term among the triples converted by the RDF triple conversion unit 300. The term terminology processing unit 400 stores in the RDF triple repository 500 a triple that does not include a major term among the triples converted by the RDF triple conversion unit 300.

The term terminology processing unit 400 compares the preceding triple stored in the RDF triple repository 500 with the term triple containing the upper term. The vocabulary processing unit 400 specifies a leading word corresponding to a vocabulary based on a leading-word specifying condition. A predicate-specific condition is a condition for judging the similarity between triples. The type of a large term, the vocabulary contained in a large term, the qualities of a vocabulary contained in a large term, the vocabulary included in a triple, and the qualities of a vocabulary included in a triple are compared . For example, the term terminology processing unit 400 may search for a preceding triple that includes a vocabulary contained in a large term. Or vocabulary processing unit 400 finds a preceding triple that contains the same vocabulary / qualities as the subject of the large term triple, the predicate kit, and the object. Where the vocabulary triple refers to the RDF triple that contains terms relative to the subject or object. The qualities include information indicating various characteristics of the vocabulary such as gender qualities such as sex, number, personality, and semantic qualities of people, objects, places, foods, organizations and the like.

The vocabulary processing unit 400 generates the restored triple by replacing the vocabulary with the leading word. The RDF triple restoring unit 450 stores the restored triples in the RDF triple repository 500.

2 is a block diagram of a vocabulary processing unit according to an embodiment of the present invention.

2, the vocabulary processing unit 400 includes a vocabulary extracting unit 410, a leading character identifying unit 430, and an RDF triple restoring unit 450.

The term extraction unit 410 determines whether a term exists in the subject and the object position. The term extraction unit 410 extracts a directive, a directive + noun, a common noun, and a young unit term display information, which are located in the subject and the object position, in large terms. Herein, the Young-Daejun term display information is information indicating that the corresponding place is empty, and may be a variable such as "? X" or "? Y ".

The term extraction unit 410 stores in the RDF triple repository 500 a triple that does not include a major term among the triples converted by the RDF triple conversion unit 300.

The leading-word specifying unit 430 compares the preceding triple stored in the RDF triple repository 500 with the corresponding term triple including the large term. A leading triple is a triple of sentences preceded by a sentence of a major term triple, possibly a triple containing a leading term.

The leading-word specifying unit 430 finds a leading triple having the same vocabulary / qualities as the large-term triple based on the leading-word-specific condition. The leading-line specifying unit 430 specifies a leading term corresponding to the large term in the found leading triple. The leading-word specifying unit 430 can apply the leading-line-specific condition according to the type of the large term, the vocabulary included in the large term, the qualities of the vocabulary included in the triple, the similarity of triples, and the like.

The RDF triple restoring unit 450 generates a restored triple by replacing a large term with a leading word. The RDF triple restoring unit 450 stores the restored triples in the RDF triple repository 500.

Next, the operation of the term terminology processing unit 400 will be described by way of example.

The first embodiment is as follows.

The RDF triple conversion unit 300 converts each of the sentences into an RDF triple as shown in Table 1 by using the SNS dashboard technology developed by SensoLo.

Subject Freddie Kit Object Sentence 1 Sen Soloji Development SNS Dashboard Technology Sentence 2 This technology Big Value of use

The term extraction unit 410 extracts "this technique" in the form of "directive + noun "

The leading-line specifying unit 430 specifies a leading term corresponding to a large term based on a leading-end specifying condition. The leading-word specifying unit 430 finds an object that morphologically includes the "noun" vocabulary included in the vocabulary in the preceding triples when the vocabulary is in the form of "directive + noun ". The leading-line specifying unit 430 specifies an object that morphologically includes a "noun" vocabulary as a leading line. That is, the leading-word specifying unit 430 specifies the object "SNS dashboard technique" of sentence 1 that includes the vocabulary "technique" included in "this technique" as a leading word.

The RDF triple restoring unit 450 restores the triple of sentence 2 as shown in Table 2.

Subject Freddie Kit Object Sentence 2 This technology Big Value of use Restored sentence 2 SNS Dashboard Technology Big Value of use

The second embodiment is as follows.

The RDF triple conversion unit 300 converts each of the sentences into "RDF triple" as shown in Table 3, by "SenSoloro developed RDF conversion technology.

Subject Freddie Kit Object Sentence 3 Sen Soloji Development RDF conversion technology Sentence 4 This company secure Related patents

The term extraction unit 410 extracts " this company "in the form of" directive + noun "

The leading-word specifying unit 430 finds an object that morphologically includes a "noun" vocabulary included in a vocabulary among preceding triples when the vocabulary is in the form of "directive + noun ". If there is no triple that morphologically includes the "noun" vocabulary included in the vocabulary, the leading-word specifying unit 430 finds an object having the same qualities as the "noun" vocabulary included in the vocabulary among preceding triples . At this time, the leading-line specifying unit 430 finds a similar-quality subject by tracing from a preceding triple nearest to a large-term triple. That is, since there is no triple in the preceding triple that morphologically includes the word "company" included in "the company " in the preceding triple, the subject" Is specified as a leading character. Here, "company" and "sen solo" share the qualities of "organization".

The RDF triple restoring unit 450 restores the triple of the sentence 4 as shown in Table 4.

Subject Freddie Kit Object Sentence 4 This company secure Related patents Restored Sentence 4 Sen Soloji secure Related patents

The third embodiment is as follows.

The RDF triple conversion unit 300 converts each of the sentences "Do healing camp on Monday, do incarnation on Tuesday, and paired on Wednesday, and record on Monday."

Subject Freddie Kit Object Sentence 5-1 healing camp Do Monday Sentence 5-2 avatar Do Tuesday Sentence 5-3 match Do Wednesday Sentence 6-1 roughness Do Monday Sentence 6-2 ? x recording roughness

The large term extraction unit 410 extracts the "pronoun" type "ger" from the triple sentence 6-1 in a large term.

The leading-line specifying unit 430 finds the leading triple in the leading triple that is the same as the triple in the sentence 6-1 including the large term. The leading-line specifying unit 430 finds the leading triple of the subject "gig, " so that the predicate kit and the object find the same leading triple. The leading-line specifying unit 430 specifies the subject of the leading triple in which the object and the predicate kit are the same as the leading line. In other words, the leading character specification unit 430 finds the triple of the sentence 5-1 including the predicate kit "do", the object "Monday" of the sentence 6-1 including the large term " Identify "healing camp" as a leading word.

The RDF triple restoring unit 450 restores the triple of the sentence 4 as shown in Table 6.

Subject Freddie Kit Object Sentence 6-1 roughness Do Monday Restored sentence 6-1 healing camp Do Monday

The fourth embodiment is as follows.

The RDF triple conversion unit 300 converts each sentence of "Listen to Jazz yesterday, listen to music " to RDF triple as shown in Table 7.

Subject Freddie Kit Object Sentence 7-1 ? x time yesterday Sentence 7-2 ? x Listen jazz Sentence 8-1 ? x Listen music Sentence 8-2 ? x Think she

The large term extraction unit 410 extracts the common noun "music" in a triple from the sentence 8-1.

The leading-word specifying unit 430 finds the leading triple in the preceding triple that is the same as the triple in the sentence 8-1 including the large term. Since the leading character specification unit 430 must find the leading character of the object "music ", the subject and the predicate kit find the same leading triple. Then, the leading-line specifying unit 430 identifies the subject and the predicate kit as an anterior-lingual object of the same leading triple. That is, the leading character specification unit 430 finds the triple of the sentence 7-2 including the predicate kit " listen "of the sentence 8-1 including the major term" music " .

The RDF triple restoring unit 450 restores the triple of the sentence 8-1 as shown in Table 8. [

Subject Freddie Kit Object Sentence 8-1 ? x Listen music Restored sentence 8-1 ? x Listen jazz

The fifth embodiment is as follows.

The RDF triple conversion unit 300 converts each of the phrases "Bob likes Americano and likes Lee" into an RDF triple as shown in Table 9. [ If there is no vocabulary corresponding to any triple, the RDF triple conversion unit 300 inputs the Young-type word term display information indicating that it is vacant.

Subject Freddie Kit Object Sentence 9 withdrawal like Americano Sentence 10 Young Hee like ? y

The term extraction unit 410 extracts the object "? Y" of the sentence 10 as a large term based on the Young-Dae term display information described in the object of the sentence 10.

The leading-line specifying unit 430 finds the leading triple in the preceding triple that is the same as the triple of the sentence 10 including the large term. Since the leading character specification unit 430 must find the leading character of the object, the subject and the predicate kit look for the same leading triple. Then, the leading-line identifying unit 430 identifies an object of the found leading triple as a leading line. That is, the leading-word specifying unit 430 finds the triple of the sentence 9 including the Freddie kit "like" in the sentence 10 and specifies the object "Americano"

The RDF triple restoring unit 450 restores the triple of the sentence 10 as shown in Table 10.

Subject Freddie Kit Object Sentence 10 Young Hee like ? y Restored Sentence 10 Young Hee like Americano

The sixth embodiment will be described.

The RDF triple conversion unit 300 converts each of the phrases "Chil Soo likes Americano " and " likes Cafe Latte" to RDF triples as shown in Table 11.

Subject Freddie Kit Object Sentence 11 withdrawal like Americano Sentence 12 ? x like Cafe latte

The term extraction unit 410 extracts the subject "? X" in the sentence 12 on the basis of the Young-Dae term display information described in the subject of the sentence 12 as a large term.

Since the leading-language specifying unit 430 must find the subject's leading word, the triple of the sentence 12 and the object and the predicate kit find the same leading triple. If there is no identical preceding triple, the leading-word specifying unit 430 finds a preceding triple including the same qualities. The leading-line specifying unit 430 specifies a subject of the found leading triple as a leading line. That is, the leading-word specifying unit 430 finds a triple of the sentence 11 including the Freddie kit "like" in the sentence 12 and includes "Americano" Specify the subject "withdraw" as a leading character.

The RDF triple restoring unit 450 restores the triple of the sentence 12 as shown in Table 12.

Subject Freddie Kit Object Sentence 12 ? x like Cafe latte Restored sentence 12 withdrawal like Cafe latte

The seventh embodiment will be described.

The RDF triple conversion unit 300 converts each sentence of "When is the going to Seoul? Tomorrow?" Into an RDF triple as shown in Table 13.

Subject Freddie Kit Object Sentence 13-1 withdrawal go When (Time) Sentence 13-2 withdrawal go Seoul (Place) Sentence 14 ? x go Tomorrow (Time)

The term extraction unit 410 extracts the subject "? X" in the sentence 14 as a large term based on the Young-Dae term display information described in the subject of the sentence 14.

Since the leading character specification unit 430 must find the subject's leading character, the triple in the sentence 14 and the object and the predicate kit look for a preceding triple having the same or similar qualities. The leading-line specifying unit 430 specifies a subject of the found leading triple as a leading line. That is, the leading-word specifying unit 430 includes a predicate kit "GADA " of the sentence 14 including the Young's word term, and the sentence 13 including the " when" -1 and specifies the subject "withdraw" in sentence 13-1 as a leading character.

The RDF triple restoring unit 450 restores the triple of the sentence 14 as shown in Table 14. [

Subject Freddie Kit Object Sentence 14 ? x go Tomorrow (Time) Restored sentence 14 withdrawal go Tomorrow (Time)

3 is a flowchart of a method for restoring a large term according to an embodiment of the present invention.

Referring to FIG. 3, the apparatus 100 restores the input text composed of text (S110).

The vocabulary restoration apparatus 100 analyzes the sentence components of the input query (S120).

The vocabulary restoration apparatus 100 converts each sentence into an RDF triple based on the sentence component of the input query (S130).

The term restoration apparatus 100 searches for a term term triple including a plurality of terms among a plurality of triples (S140).

The term-based restoration apparatus 100 compares the preceding triples of the major term triple with the major term triple to specify a leading term corresponding to the major term (S150).

The term restoration apparatus 100 restores a major term triple by replacing the major term with a leading term (S160).

The term term recovery apparatus 100 stores the restored term term triple (S170).

4 is a flowchart of a method for restoring a large term according to another embodiment of the present invention.

Referring to FIG. 4, the vocabulary processing unit 400 receives a plurality of RDF triples converted from the input statements (S210).

The term terminology processing unit 400 determines whether each triple includes a term (S220). The term terminology processing unit 400 determines whether there is a term in the place of the subject and the object. The vocabulary processing unit 400 extracts a directive, a directive + noun, a common noun, and a young group term display information, which are located at a subject and an object position, in a large term.

The term terminology processing unit 400 stores a triple that does not include a major term in the RDF triple repository 500 (S230).

When a large term term including a large term is found, the large term processing unit 400 finds an associated preceding triple having the same vocabulary or qualities as the large term triple based on the leading term specific condition (S240). Leading word specific conditions include the form of a vocabulary, the vocabulary contained in a vocabulary, the qualities of a vocabulary contained in a triple, and the similarity of triples. For example, the vocabulary processing unit 400 finds an object that morphologically includes a "noun" vocabulary included in the vocabulary in the preceding triples when the vocabulary is in the form of "directive + noun ". If there is no triple that morphologically includes the "noun" vocabulary contained in the vocabulary, the vocabulary processing unit 400 finds an object having the same qualities as the "noun" vocabulary included in the vocabulary of the preceding triples . If the subject includes a vocabulary, the vocabulary processing unit 400 finds a predecessor triple that is the same as the predicate kit of the vocabulary triple and the vocabulary / vocabulary contained in the object. If the object includes a large term, the large term processing unit 400 finds a predecessor triple that is identical to the predicate kit of the large term triple and the lexical / vocabulary qualities contained in the subject.

The term terminology processing unit 400 specifies the leading term corresponding to the major term in the related preceding triple (S250). For example, if the large term is in the form of a "directive + noun ", it may be used to formally include the" noun "vocabulary in the preceding triples, Specify. If there is no subject having the same vocabulary / qualities, the term terminology processing unit 400 identifies the subject in the same position as the preceding term in the preceding triple found as a leading character. That is, if a large term is in the subject position, the subject in the subject of the preceding triple can be identified by the leading character, and if the large term is in the place of the object, the subject in the object of the preceding triple can be identified by the leading character.

The vocabulary processing unit 400 replaces a vocabulary with a leading word to restore a vocabulary triple (S260).

The term terminology processing unit 400 stores the restored term term triple in the RDF triple repository (S270).

As described above, according to the embodiment of the present invention, the meaning of the sentence and the context information are represented by the RDF triple, so that the leading word can be accurately specified as compared with the method of determining the leading word based on the degree of correspondence between the individual grammar qualities and the semantic qualities. According to the embodiment of the present invention, it is possible to precisely specify a leading term corresponding to a large term even if the candidate of the leading term having similar grammatical qualities such as sex, number, and personality and the semantic qualities of people, objects, places, have. Also, according to the embodiment of the present invention, it is possible to precisely specify the leading word corresponding to the Young-Dang term. Therefore, according to the embodiment of the present invention, a better performance can be expected for automatic language processing apparatuses.

The embodiments of the present invention described above are not implemented only by the apparatus and method, but may be implemented through a program for realizing the function corresponding to the configuration of the embodiment of the present invention or a recording medium on which the program is recorded.

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, It belongs to the scope of right.

Claims (19)

Resource Description Framework (RDF) A triple-based large term restoration device,
An RDF triple conversion unit for converting each sentence included in the input statement into an RDF triple, and
Wherein the RDF triple conversion unit searches for a large term term triple including a large term among a plurality of triples converted by the RDF triple conversion unit and compares the large term triple with a preceding triple of the large term triple to specify a leading term of the large term,
Lt; / RTI >
The term terminology processing section
Finding at least one related preceding triple having the same vocabulary or the same qualities as the corresponding term triple among the preceding triples,
Determining at least one of a subject, a predicate kit, and an object of the large term triple and at least one of the subject, the predicate kit, and the object of the at least one related preceding triple, Wherein the at least one preceding leading triple identifies the leading term of the at least one terminology. The method of claim 1,
The term terminology processing section
And a triple term having a term in a place of the object and an object among the plurality of triples is extracted as a large term triple. The method of claim 1,
The term terminology processing section
A terminology restoration device that judges, based on a vocabulary, a vocabulary containing a directive, a normal noun, a pronoun, and a Young's term display information indicating that the subject or object of the RDF triple is empty. delete The method of claim 1,
The term terminology processing section
And wherein the vocabulary described in the found subject or object is specified as a preceding word when the vocabulary is a form including an instruction and a noun, the subject or object including the noun in the preceding triples. The method of claim 5,
The term terminology processing section
Wherein the subject or object including the noun is not found among the preceding triples, the subject or object containing the same vocabulary with the noun is searched. The method of claim 1,
The term terminology processing section
And wherein if the vocabulary is located at a subject position, the object of the preceding triple and the object of the vocabulary triple are compared with the predicated kit to find the related preceding triple and the subject of the related preceding triple is specified as the preceding word. 8. The method of claim 7,
The term terminology processing section
And wherein if the large term is located in an object position, comparing the subject of the preceding triple with the subject term triple and the predicate kit to find the related preceding triple and specify the object of the related preceding triple as a leading character. The method of claim 1,
The term terminology processing section
Term restorer by replacing the large term with a leading character, restoring the large term triple, and storing the restored large term triple. The method of claim 1,
The term terminology processing section
A large term extraction unit for extracting a large term triple including a large term from the plurality of triples converted by the RDF triple conversion unit,
A preceding-word specifying unit for specifying a preceding term of the vocabulary by comparing the preceding-term triple with the preceding-term triple of the vocabulary triple, and
An RDF triple restoring unit for restoring the large term triple by replacing the large term with a leading term,
The term restoration device comprising: A method for restoring a large term based on a Resource Description Framework (RDF) triple,
Receiving a plurality of RDF triples corresponding to each sentence of an input query;
Searching for a large term triple including a large term among a plurality of input triples,
Finding at least one related preceding triple having the same vocabulary or qualities as the corresponding term triple among the preceding triples of the corresponding term triple, and
Determining at least one of a subject, a predicate kit and an object of the corresponding term triple and at least one of the subject, the predicate kit and the object of the at least one related preceding triple, Identifying the leading term of the vocabulary in at least one related preceding triple
/ RTI > 12. The method of claim 11,
The step of finding the term triple
When at least one of a vocabulary including a directive, a common noun, a pronoun, and a Young-tongue term display information is described in at least one of a subject and an object of an arbitrary triple,
Wherein the Young's Unit term display information is information indicating that a subject or object of the corresponding triple is empty. 12. The method of claim 11,
The step of finding the related preceding triple
If the large term is a form including a directive and a noun, searching for a subject or object containing the noun from preceding triples,
The step of extracting the lead-
A vocabulary restoration method for extracting a vocabulary described in a found subject or object with a leading term. The method of claim 13,
The step of finding the related preceding triple
If the subject or object including the noun is not found among the preceding triples, searching for the subject or object having the same qualities as the noun. 12. The method of claim 11,
The step of finding the related preceding triple
Comparing the preceding triples and the object of the corresponding term triple with the predicated kit to find a related preceding triple having the same vocabulary or qualities as the corresponding term triple when the corresponding term is located at the subject position. 16. The method of claim 15,
The step of identifying the preceding term
And identifying the subject of the related preceding triple as a leading word. 12. The method of claim 11,
The step of finding the related preceding triple
Comparing the preceding triples with the subject of the corresponding term triple and the predicated kit to find an associated preceding triple having the same vocabulary or qualities as the corresponding term triple. The method of claim 17,
The step of identifying the preceding term
And the object of the related preceding triple is specified as a leading word. 12. The method of claim 11,
Replacing the large term with a leading term to restore the large term triple; and
The step of storing the restored term term triple
Further comprising the steps of:

Images