WO2017094967A1 - Natural language processing schema and method and system for establishing knowledge database therefor - Google Patents

Abstract

Provided are a method and a system for establishing a natural language processing schema for a text on the basis of identification information including location information related to each of identified words and relationship information indicating the dependency relationship between words included in the text by identifying the words included in the text, associating the identification information with each of the identified words, and identifying the relationship information.

Description

Natural language processing schema and its knowledge database construction method and system

The following embodiments relate to a method for building a natural language processing schema and its knowledge database, and in particular, a method for constructing a schema about natural language processing annotation information for each word and its knowledge database by individualizing words in the text. Is associated with.

Recently, as research on the semantic web and big data is active, the demand for processing of language resources is rapidly increasing, and various studies for constructing language resources based on ontology are being conducted.

On the other hand, in order to solve the problem related to the accessibility to the natural language processing annotation (NLP annotation) that exists in various forms of data, researches are attempting to define the natural language processing annotation information based on the ontology.

Natural language processing The existing method for processing annotation information uses a method of ontotizing metadata of text and simply storing annotation information in sentence units. This method has the advantage of allowing the user to use the same type of data, but there is a problem in that the search quality and speed are deteriorated when the user wants to obtain detailed information about a word or a detailed structure between words. .

Therefore, there is a need for a method of processing natural language processing annotation information that provides a user with stable and reliable information while increasing the user's access to natural language processing results.

Korean Patent Registration No. 10-1476225 (Registration Date December 18, 2014) receives combination data consisting of a combination of natural language and formula, separates natural language and formula, respectively, and constructs each of the separated natural language and formula. Apparatus and methods are disclosed for enabling the indexing of natural language and expressions based on analyzing information.

The information described above is merely for the sake of understanding, and may include content that does not form part of the prior art, and may not include what the prior art may suggest to those skilled in the art.

It is possible to provide a method and system for associating annotation information associated with each word for each word included in the text, and building a natural language processing schema for the text.

Index of a string such as a word can provide a method and system for constructing a linguistic knowledge database that allows the usage of the word, the structure of the sentence associated with the word, or the linguistic structure and semantic relationship between multiple words. .

In one aspect, identifying at least one word included in text, associating identification information with location information associated with the identified word to the identified word, the identification based on the identification information Identifying relationship information between the word and other words included in the text, and constructing a schema for the text based on the identification information and the relationship information. Is provided.

The words contained within the text may be characters or continuous strings in the text that have meaning by themselves.

Identifying the at least one word may include determining target words valid for building the schema among words included in the text and extracting the determined target words.

The associating the identification information may associate the identification information of each of the extracted target words with a corresponding target word.

The identifying of the relationship information may identify relationship information between at least two target words among the extracted target words.

The constructing of the schema may include determining each of the extracted target words and a plurality of nodes associated with each of the identification information associated with the extracted target words, based on the identified relationship information. Connecting a first node associated with a first target word and a second node associated with a second target word associated with the first target word, and a node representing identification information of the first node and the first target word; Connecting the second node and a node representing identification information of the second target word.

The first target word may be a word included in a search word input by a user.

Information regarding an association between the identified word and the identification information may be stored in a database.

The identification information may include at least one of Uniform Resource Identifier (URI) information of the identified word, information indicating a part-of-speech of the identified word, and linguistic information of the identified word. .

The identification information is information indicating a position of the text in a document containing the text, information indicating a position in the text of the identified word, information indicating a stemming result of the identified word, the identified word It may include at least one of information indicating a part-of-speech of the information, information indicating the entity name of the identified word, and information indicating a tag of the entity name of the identified word. The identifying information may include natural language processing annotation information and other information about the text.

If the identified word is a word that is also included in other text different from the text, identifying the relationship information may identify relationship information between the identified word and other words included in the other text.

The constructing of the schema may build a schema for the other text based on the identification information of the identified word and relationship information between the identified word and other words included in the other text.

The relationship information may include information indicating a dependency relationship between the identified word and the other word.

The dependency relationship may be at least one of a spelling relationship, a subject-object relationship, and a mathematical relationship between the identified word and the other word.

In another aspect, identifying at least one word included in text, associating identification information including location information associated with the identified word to the identified word, based on the identification information; Identifying relationship information between the identified word and other words included in the text, building a schema for the text based on the identification information and the relationship information, and storing the constructed schema in a database. A method of building a database of schemas for text is included.

In another aspect, identifying at least one word included in text, associating identification information including location information associated with the identified word to the identified word, based on the identification information Identifying relationship information between the identified word and other words included in the text, constructing a schema for the text based on the identification information and the relationship information, receiving a search word from a user, and the search word If includes the identified word, providing a search result based on the established schema as a search result for the user's search request by the search term, is provided.

In yet another aspect, identifying at least one word included in text, associating identification information including location information associated with the identified word to the identified word, and identifying the based on the identification information. A control unit for identifying relationship information between the identified word and another word included in the text, and constructing a schema for the text based on the identification information and the relationship information, and the constructed schema and the identified word and the identification information. Provided is a system for processing information included in text, comprising a storage unit for storing at least one of information relating to an association of a.

By constructing a natural language processing schema for text by individualizing each word contained in the text, and building a database, a method and system for providing stable and reliable information to the user while increasing the accessibility of the user's information is provided. Is provided.

Obtain information about the usage of the word, the structure of sentences associated with the word, or the linguistic structure and semantic relationships between the words, by indexing strings such as words within the schema, without the use of additional external indexing devices or additional analysis. A method and system are provided to enable this.

1A-1D illustrate a method of building a knowledge structure corresponding to a representation in the real world for text, according to an example.

2 illustrates a method of building a natural language processing schema and knowledge database for text, according to an embodiment.

3 illustrates a system for building a natural language processing schema for text and building its database, according to one embodiment.

4 is a flow diagram illustrating a method of building a natural language processing schema for text and constructing a database according to an embodiment.

5 is a flowchart illustrating a method of identifying word (s) included in text according to an example.

6 is a flow diagram illustrating a method of building a natural language processing schema for text, according to an example.

7 is a flowchart illustrating a method of providing a search result based on a natural language processing schema for text according to an embodiment.

8 illustrates a natural language processing schema for text, according to an example.

9A and 9B illustrate a natural language processing schema for a plurality of texts, according to an example.

Hereinafter, exemplary embodiments will be described in detail with reference to the accompanying drawings. Like reference numerals in the drawings denote like elements.

1A-1D illustrate a method of building a knowledge structure corresponding to a representation in the real world for text, according to an example.

In Figures 1A-1D, the text "Ford was born in the United States on July 14, 1913. He began his presidency in 1977. ..." in the real world representation of the text and The knowledge structure built in the text environment is shown.

The representation of the text in the real world may indicate how the information contained in the text is perceived in the real world. For example, in the real world representation of the text, "Gerald Ford" was born in the United States (born in) on July 14, 1913 (born at), the 38th President of the United States, and in August 1974. The term may begin on the 9th, ending on January 20, 1977, and died at December 26, 2006 in the United States.

The knowledge structure of the text may represent the information recognized in the representation of the text in the real world in the text environment. For example, the knowledge structure of text may be a natural language processing schema for text stored on an ontology knowledge base (knowledge database). For example, the knowledge structure of the text is "gerald pod" (res: gerald_ford) was born in the United States (prop: birthPlace), was born on July 14, 1913 (prop: birthDate), and the president of the United States ( prop: job), starting on August 9, 1974 (prop: startYear), ending on January 20, 1977 (prop: endYear), and died on December 26, 2006 (prop: deathYear Can be indicated. The knowledge structure of the text can be built along the time axis and can be constructed as close as possible to the information recognized in the representation in the real world (or to include the maximum information recognized in the representation in the real world).

The knowledge structure of the text may be constructed by the Natural Language Processing Annotation Knowledge Base Format (NKF). For example, as shown in FIGS. 1B-1D, each of the words (objects) contained in the text and / or information associated with the text may be individualized and linked as a node, and the relationship between the words may be extracted to indicate information representing that relationship. By being associated with a node as an annotation, a knowledge structure for that text can be built. NKF may represent the desired natural language processing annotation base.

In FIGS. 1A-1D, "res" may represent a knowledge entity (eg, a noun or a word) in text, and a connection (relationship extraction) between entities may be made based on relationship information between entities. For example, when an entity is a verb, relationship information with at least two other entities may be required for extracting a relationship with another entity of the entity.

The NKF may build a knowledge structure that includes annotation information based on linguistic information analysis for the text, and the constructed knowledge structure may be built as a database. In addition, according to a user's search request, search information associated with text and knowledge structures may be provided as a search result.

A knowledge structure (schema) for text and a method of constructing the database will be described in more detail with reference to FIGS. 2 to 9B to be described later.

2 illustrates a method of building a natural language processing schema and knowledge database for text, according to an embodiment.

In FIG. 2, a natural language processing schema including annotation information related to text is constructed through natural language processing on (input) text according to the NKF described above with reference to FIG. 1, and the knowledge base (database) The method of construction is shown.

In step 210, the natural language processing annotation tool may perform natural language processing on the text (sentence) (s) manually or automatically. In operation 220, the natural language processing annotation unit may store and output a processing result by the natural language processing annotation tool, that is, an annotation result (comment information) for each text (statement). The output result of the natural language processing annotation unit may be different depending on the processing result in step 210. In step 230, the word extractor may extract the target word (s) to be used in constructing the schema from the text (sentence), depending on the desired purpose. For example, a noun or a verb included in a sentence may be extracted as a target word. In step 240, the target words extracted by the word extractor may be individualized and indexed. The individualization of the target word may be performed within, for example, a Resource Description Framework (RDF) schema. In step 250, additional information on the extracted target words (identification information about each target word and / or relationship information between the target words) may be obtained from the natural language processing annotation unit and associated with the corresponding target word. have. In step 260, the binomial relationship between the extracted target words may be obtained from the natural language processing annotation unit and associated with the target word, and the target words may be connected to each other according to the obtained binomial relationship. In step 270, information representing an association between the graph (corresponding to the schema for the text) and / or the extracted word and its annotation information constructed by the steps 210 to 260 described above may be stored in the knowledge database. Can be.

The above described steps 210 to 270 may be performed by the system 300 or a configuration thereof to be described later.

A knowledge structure (schema) for text and a method for constructing the database thereof will be described in more detail with reference to FIGS. 3 to 9B to be described later.

Since the technical contents described above with reference to FIG. 1 may be applied as it is, a more detailed description will be omitted below.

3 illustrates a system for building a natural language processing schema for text and building its database, according to one embodiment.

The illustrated system 300 may correspond to a system for building a natural language processing schema for text through natural language processing for text according to the NKF described above with reference to FIGS. 1 and 2.

The system 300 may include a controller 310, a communicator 320, and a storage 330. The controller 310 may manage components of the system 300 and execute a program or application used by the system 300. For example, the controller 310 may execute a program or an application for performing natural language processing of the (input) text and building a schema for the text. In addition, the controller 310 may process operations necessary for executing a program or an application and processing data. The controller 310 may be at least one processor of the system 300 or at least one core within the processor.

The communicator 320 may be a device for communicating with a different device (s) or server than the system 300. For example, the communicator 320 may receive text (or a search term) from other device (s) or a server. Although not shown, the communication unit 320 may include one or more antennas for transmitting and receiving signals and information with other device (s) or a server. The communicator 320 may be a hardware module such as a network interface card, a network interface chip and a networking interface port of the system 300, or a software module such as a network device driver or a networking program.

The storage unit 330 may store information related to a program or application for constructing a schema executed by the control unit 310 and / or a built schema. The storage unit 330 may be a database that stores an association relationship between a word included in text or a corresponding word and identification information of the word. Alternatively, the storage unit 330 may be a database that stores the built schema. Unlike shown, storage 330 may be configured as a separate device from system 300.

The knowledge structure (schema) for the text and how the database is built by the system 300 is described in more detail with reference to FIGS. 4-9B described below.

Since the technical contents described above with reference to FIGS. 1 and 2 may be applied as it is, a more detailed description will be omitted below.

4 is a flow diagram illustrating a method of building a natural language processing schema for text and constructing a database according to an embodiment.

In operation 410, the controller 310 may identify at least one word included in the text. The text may be a sentence including at least one word present in the document. A word contained within the text may be a character or a continuous string in the text that has meaning in itself. For example, a word may be a noun (a phrase), a verb (a phrase), an adjective (a phrase) or an adverb (a phrase). Or the word can be a survey.

In operation 420, the controller 310 may associate identification information including location information related to the word identified for the word identified in operation 410. The identification information may be generated by the controller 310 through a natural language processing task for the text. Identification information associated with a word may be at least one of Uniform Resource Identifier (URI) information of the identified word, information representing a part-of-speech of the identified word, and linguistic information of the identified word. The URI information of the identified word is information about which document the identified word is in (document position information), information about which sentence is in which document (position position information), and at which position within the sentence. It may include at least one of information about whether. For example, the URI information of the identified word may include URL information of the document in which the word exists.

In addition, the identification information includes information indicating the position of the text (sentence) in the document containing the text (sentence), information indicating the position in the text (sentence) of the identified word, and information indicating the stemming result of the identified word. , Information indicating the part-of-speech of the identified word, information indicating the entity name of the identified word, and information indicating a tag of the entity name of the identified word. The entity name may be a category classified based on the meaning of the word. For example, the individual name may represent a person's name, organization name or place name. The entity name may correspond to "CLASS" in FIGS. 1A-1D. The part of speech of a word may be, for example, a verb (a phrase), a noun (a phrase), an adjective (a phrase) or an adverb (a phrase).

The identification information may be included in annotation information (additional information) obtained for the text by the natural language processing of the steps 210 and 220 described above with reference to FIG. 2.

In operation 430, the controller 310 may identify relationship information between the identified word and other words included in the text based on the identification information. The relationship information may be information indicating a dependency relationship between the identified word and other words in the text. The relationship information may be generated by the controller 310 through a natural language processing task for text. The dependency relationship may be, for example, at least one of a magical relationship between the identified word and another word, a subject-object relationship, a mathematical relationship, an adverb relationship (to determine whether one is an adverb), and other dependency relationships.

The relationship information may be included in annotation information (additional information) obtained for text by the natural language processing of steps 210 and 220 described above with reference to FIG. 2.

In operation 450, the controller 310 may build a schema for the text based on the identification information and the relationship information of the words in the text. Constructed Schema (NKF Schema) may be an RDF Schema that is constructed by individualizing identified words in text. In other words, the constructed NKF schema can express information about each word in the text using an RDF graph. The schema and the method of constructing the schema are described in more detail with reference to FIGS. 6, 8, 9a and 9b which will be described later.

If the word identified in the text is a word that is also included in other text (s) different from the text, that is, if the same word is present in a sentence or sentences in a plurality of documents, then at steps 420 and 430, The controller 310 generates identification information for each word in association with not only the text but also other text (s), associates with each word, and relationship information between each word and other words included in the other text (s). Can be identified.

In other words, in step 430, when the identified word is also included in the second text, which is another text different from the first text, the control unit 310 indicates relationship information between the identified word and the other words included in the second text. In operation 440, a schema for the second text may be established based on the identification information of the identified word and relationship information between the identified word and other words included in the second text. . For example, a natural language processing schema can be built for all texts containing the same word. A natural language processing schema built for a plurality of texts is described in more detail with reference to FIGS. 9A and 9B.

In operation 450, the controller 310 may store the schema established in operation 440 in the database (or the storage 330). As such, a knowledge database about the schema can be built.

Meanwhile, the controller 310 may store information regarding an association relationship between the identified word in the text and the identification information of the identified word in the database (or the storage 330). In other words, the storage unit 330 is a built schema; And information about an association between the identified word and the identification information. For example, the controller 310 may store an association relationship between each of the individualized words of text and identification information thereof in the database (or the storage 330). Information about an individualized word is stored in a database, so that the individualized word can have its own unique value, which can speed up the search for that word by indexing the individualized word and build an RDF graph around the word. You can build the schema by creating it.

Since the technical contents described above with reference to FIGS. 1 to 3 may be applied as it is, a more detailed description will be omitted below.

5 is a flowchart illustrating a method of identifying word (s) included in text according to an example.

In FIG. 5, a method of identifying (extracting) word (s) from text is described in more detail. Steps 510 and 520 to be described below may be included in step 410 described above with reference to FIG. 4.

In operation 510, the controller 310 may determine target words valid for constructing a schema among words included in the text. The "target word" may be a string having meaning in itself and may correspond to a word in the text constituting the node in constructing the schema (Goal of String; GoS). For example, determination of target words may be performed by classifying words included in the text according to the word of speech or a predetermined set of tags, or classifying words included in the text according to other morphological analysis. "Target word" is each word in the text that is not a survey and may be, for example, a verb or a noun.

In operation 520, the controller 310 may extract target words determined in operation 510. The extracted target words may be used to determine the node of the schema to be built in step 440.

The method of building the schema is signed in more detail with reference to FIG. Since the technical contents described above with reference to FIGS. 1 to 4 may be applied as it is, a more detailed description will be omitted below.

6 is a flow diagram illustrating a method of building a natural language processing schema for text, according to an example.

Steps 610 to 630 to be described below may be included in step 440 described above with reference to FIG. 4.

In operation 420, the controller 310 may associate identification information of each of the target words extracted in operation 510 described above with reference to FIG. 5, with a corresponding target word. In operation 430, the controller 310 may identify relationship information between at least two target words among the extracted target words.

In operation 610, the controller 310 may determine each of the extracted target words and a plurality of nodes associated with each of the identification information associated with the extracted target words. For example, the controller 310 may determine nodes associated with each of the extracted target words, and may further determine nodes associated with information about each target word indicated by the identification information.

In step 620, the controller 310 based on the relationship information identified in step 430, the first node associated with the first target word among the extracted target words and the second target word associated with the first target word. Connect a second node associated with the. The second target word may have a dependency relationship with the first target word. In the connection between nodes, information representing the relationship information may be added and included in the schema.

In operation 630, the controller 310 may connect a node representing identification information of the first target word with a first node, and connect the node representing identification information of the second target word with the second node. . In the connection between nodes, information representing identification information may be added and included in the schema.

The constructed schema may be a graph having a tree structure in which the first node exists above. The constructed schema is described in more detail with reference to FIGS. 8, 9A, and 9B.

The first target word associated with the first node may be a word included in a search word input by a user. For example, the first target word may be a word to be indexed included in a search word input by a user.

Since the technical contents described above with reference to FIGS. 1 to 5 may be applied as it is, a more detailed description will be omitted below.

7 is a flowchart illustrating a method of providing a search result based on a natural language processing schema for text according to an embodiment.

In operation 710, the controller 310 may receive a search word from a user through the communication unit 320.

In operation 720, if the received search word includes a word (corresponding to the identified word described above with reference to FIGS. 4 to 6) associated with the constructed schema, the controller 310 generates a search result based on the constructed schema. It may be provided as a search result for the user's search request by the search word. For example, information related to the associated word included in the built schema may be provided as a search result. That is, a search result including abundant information can be provided to the user at high speed without the help of an external indexing device.

Since the technical contents described above with reference to FIGS. 1 to 6 may be applied as they are, a detailed description thereof will be omitted below.

8 illustrates a natural language processing schema for text, according to an example.

The graph shown in FIG. 8 may correspond to an example of an NKF graph included in a schema built by the system 300 described above with reference to FIGS. 4 through 7.

In FIG. 8 an NKF graph is shown for the text (phrase) Ford was born in the United States on July 14, 1913.

The lbox may be a database (storage unit 330) that stores an association relationship between a word included in text or a corresponding word and identification information of the word. Information about each word can be individualized and stored in an lbox with a unique value. In other words, lbox can be a database configuration of a built-in text schema.

In the following, the NKF ontology terminology and annotation format used in the depicted NKF graph are described. The annotation used may correspond to information included in the identification information and the relationship information for each of the words described above. The annotation format described below is an example, and the annotation format used in the embodiment may be different from the following.

1. Formatting comments for words

1) nif: StringURI

The String URI may be a unique value representing location information of a string of words. String URIs may be expressed as follows according to the NIF2.0 standard and the RFC 5147 standard.

<URL # charx, y>

x may be the start position of the word string, y may be the end position. At this time, the location information starts from 0, and the spacing may not be considered. The location information shall refer to the following in conformity with the NIF2.0 standard.

context (6.3.1.1), begin index (6.3.1.2), end index (6.3.1.3)

2) nif: referenceContext

The referenceContext may indicate the entire position of the sentence where the word came from. That is, it can represent the entire position of the sentence from which the word appeared in the knowledge base. The referenceContext can be useful when you want to identify the context of the word.

The ontology type may be "ObjectProperty", the domain may be "nif: String-entity" and the range may be "nif: StringURI".

3) nif: beginIndex

beginIndex may indicate the first position of a word in the knowledge base.

The ontology type may be "DatatypeProperty", the domain may be "nif: String-entity", and the range may be "xsd: nonNegativeInger".

4) nif: endIndex

endIndex may indicate the last position of a word in the knowledge base.

The ontology type may be "DatatypeProperty", the domain may be "nif: String-entity", and the range may be "xsd: nonNegativeInger".

5) nkf: String-entity

String-entity can represent an entity on the knowledge base, ie words in the NKF. The ontology type is "Class".

6) nkf: hasStringURI

hasStringURI may be a property that indicates where a word in the knowledge base is located in the actual sentence.

The ontology type may be "DatatypeProperty", the domain may be "nif: String-entity", and the range may be "nif: StringURI".

7) nkf: anchorOf

nkf: anchorOf may be a property indicating where a word in the knowledge base exists in the sentence.

The ontology type may be "DatatypeProperty" and the domain may be "nif: String-entity".

2. Part of speech information comment format

1) nif: oliaLink

oliaLink may represent a part-of-speech tag information annotation that is the result of stemming the words in the knowledge base.

The ontology type may be "ObjectProperty", the domain may be "nif: StringURI", and the range may be "Individual".

2) nif: oliaCategory

oliaCategory can represent part-of-speech concept annotations that result from stemming the words in the knowledge base.

The ontology type may be "ObjectProperty", the domain may be "nif: StringURI", and the range may be "Class".

3. Object Name Information Comment Format

The entity name information uses the NERD ontology, and the annotation processing can use the W3C ITS ontology.

1) itsrdf: taClassRef

taClassRef can represent an entity name concept annotation for words in the knowledge base.

The ontology type may be "ObjectProperty", the domain may be "nif: StringURI", and the range may be "Class".

2) itsrdf: taIdentRef

taIdentRef may represent an entity name tag comment for a word in the knowledge base. The ontology type may be "ObjectProperty", the domain may be "nif: StringURI", and the range may be "Individual".

4. Parsing Information Comment Format

The result of natural language processing analysis exists in a form in which the analysis results of the unique values of the sentences are arranged around the sentences. Therefore, in order to express this in terms of words, a new property definition is needed to express dependency relationships between words.

1) nkf: sbj

sbj indicates a dependency relationship between each String URI, and in this case, it may mean that Subject is in a subject relationship with respect to Object.

The ontology type may be "ObjectProperty", the domain may be "nif: StringURI", and the range may be "nif: StringURI".

2) nkf: obj

sbj indicates a dependency relationship between each String URI, and in this case, it may mean that Subject is in object relation to Object.

The ontology type may be "ObjectProperty", the domain may be "nif: StringURI", and the range may be "nif: StringURI".

3) nkf: ajt

ajt indicates a dependency between each String URI, and in this case, it may mean that Subject is an adverb with respect to Object.

The ontology type may be "ObjectProperty", the domain may be "nif: StringURI", and the range may be "nif: StringURI".

4) nkf: josa

josa indicates the dependency between each String URI. In this case, it can mean that Object is a subject's investigation.

The ontology type may be "ObjectProperty", the domain may be "nif: StringURI", and the range may be "nif: StringURI".

5) nkf: dp

dp can indicate other dependencies between each String URI. The ontology type may be "ObjectProperty", the domain may be "nif: StringURI", and the range may be "nif: StringURI".

By using the above-described annotation formats, it is possible to express information about words contained in the text in the form of an RDF graph as shown.

Since the technical contents described above with reference to FIGS. 1 to 7 may be applied as it is, a more detailed description will be omitted below.

9A and 9B illustrate a natural language processing schema for a plurality of texts, according to an example.

9A and 9B are included in a schema built for a plurality of texts (eg, sentence 1: "Gates co-founded Microsoft", sentence 2: "Gates has partnership with Paul") containing the common word "Gates", for example. It may correspond to an example of the NKF graph.

In the following, specific application examples will be described.

In the example below, an individualization of the words "Gates" and "Micosoft" is made in the RDF schema.

Namespace URIs must be specified.

@prefix nkf: <http://example.nkf.org/nkf/>

@prefix: <http://example.nkf.org/entity/>

Entitization of the string

: String rdf: type nkf: Entity

Example sentence: "Gates co-founded Microsoft"

: Gates rdf: type nkf: Entity

: Microsoft rdf: type nkf: Entity

An individualized string can be a "noun", a "verb", a "named entity", or any other possible string of interest to the user.

A unique value (URI) is assigned to the individualized word. In this example, for the individualized word "Gates", a unique value for each of the "Gates" of the web document is given based on the location information.

The individualized string can be mapped to the corresponding String URI and / or Sentence URI.

: String nkf: hasStringURI <StringURI>

: String nkf: hasSentenceURI <SentenceURI>

Example sentence: "Gates co-founded Microsoft"

: Gates

nkf: hasStringURI

<http://example.org/example.html#char=0,5>;

nkf: hasSentenceURI

<http://example.org/example.html#char=0,25>

For an eigenvalue (URI), map each actual word. You can map a string to the corresponding String URI and / or Sentence URI.

<StringURI> rdf: type nkf: String

<StringURI> nkf: string "STRING"

<StringURI> nkf: hasSentenceURI <SentenceURI>

Example sentence: "Gates co-founded Microsoft"

<http://example.org/example.html#char=0,5>

rdf: type nkf: String;

nkf: string "Gates";

nkf: hasSentenceURI

<http://example.org/example.html#char=0,25>

The part-of-speech information of the word is added as an annotation for the unique value of each word given above (Part-of-speech tag annotation).

<StringURI> nkf: pos <postag>

Example sentence: "Gates co-founded Microsoft"

<http://example.org/example.html#char=0,5>

nkf: pos tag: ProperNoun

The entity name tag information of the word is added as an annotation for the unique value of each word given above (NER tag annotation).

<StringURI> nkf: ner <NERtag>

Example sentence: "Gates co-founded Microsoft"

<http://example.org/example.html#char=0,5>

nkf: ner tag: Person

For the unique value of each word given above, the dependency of the phrase (word) is added as a comment.

<StringURI> nkf: dependency <StringURI>

<StringURI> nkf: role <Dependency_Role>

Example sentence: "Gates co-founded Microsoft"

<http://example.org/example.html#char=0,5>

nkf: dependency

<http://example.org/example.html#char=6,16>;

nkf: role

role: nsbj

9A and 9B, even if the same word ("Gates") is present in the plurality of texts, a different StringURI (StringURI 1, 4) is assigned to each text so that the information associated with each text is contextual. Can be distinguished according to.

In addition, as illustrated, the NKF graph for sentence 1 and the graph for sentence 2 are related to each other, so that information about the relationship between "Paul" and "Microsoft" may also be obtained. For example, if a user searches for "Paul", the information that "Paul" is co-founder of "Microsoft" may also be obtained without the help of a separate external indexing device.

In the absence of a schema for NKF indexes such as that shown in FIG. 9B, if you want to find all the cases where a common word ("Gates") is found, the number of stringURIs for all stringURIs The search speed may be slower in proportion to the number of stringURIs (ie, in the case of the conventional NIF, each stringURI should be checked for "Gates" and "Microsoft" respectively, and a nif: dependency exists. In contrast, in the present embodiment, since "Gates" and "Microsoft" are already registered in the knowledge base, nif: dependency can be searched without querying stringURI.

Since the technical contents described above with reference to FIGS. 1 to 8 may be applied as it is, a more detailed description will be omitted below.

The description of the above embodiments may be similarly applied to texts written in languages other than Korean and English.

The foregoing description may relate to the international standard of ISO / TC 37 / SC 4 Language Resource Management.

In addition, the contents described in the following standards may be applied to the description related to the above-described embodiments.

World Wide Web Consortium (W3C)

Resource Description Framework (RDF)

RDF-http://www.w3.org/RDF/

RDFS 1.1 (RDF Schema)-http://www.w3.org/TR/rdf-schema/

OWL (Web Ontology Language)

OWL-http://www.w3.org/TR/owl-features/

ITS (Internationalization Tag Set)

ITS 2.0-http://www.w3.org/TR/its20/

NIF (NLP Interchange Format)

NIF core

-http://persistence.uni-leipzig.org/nlp2rdf/ontologies/nif-core/nif-core.html

NERD (Named Entity Recognition and Disambiguation)

-http://nerd.eurecom.fr/ontology

OnLiologies of Linguistic Annotation (OLiA)

-http://nachhalt.sfb632.uni-potsdam.de/owl/

The derivation of the syntactic structure of the text, the format of its analysis, and its expression may be applied to those described in the standard "ISO 24615-1: 2014 Language resource management-Syntactic annotation framework (SynAF)-Part 1: Syntactic model". .

The term annotation may correspond to a feature-value pair denoting a linguistic property of a linguistic segment described in the standard "ISO # 24615-1: 2014 Syntactic Annotation Framework, 3.9."

Morphological analysis of the text may be referred to the content described in the standard "ISO 24611: 2012 Language resource management-Morpho-syntactic annotation framework (MAF)".

The apparatus described above may be implemented as a hardware component, a software component, and / or a combination of hardware components and software components. For example, the devices and components described in the embodiments may be, for example, processors, controllers, arithmetic logic units (ALUs), digital signal processors, microcomputers, field programmable arrays (FPAs), It may be implemented using one or more general purpose or special purpose computers, such as a programmable logic unit (PLU), microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to the execution of the software. For convenience of explanation, one processing device may be described as being used, but one of ordinary skill in the art will appreciate that the processing device includes a plurality of processing elements and / or a plurality of types of processing elements. It can be seen that it may include. For example, the processing device may include a plurality of processors or one processor and one controller. In addition, other processing configurations are possible, such as parallel processors.

The software may include a computer program, code, instructions, or a combination of one or more of the above, and configure the processing device to operate as desired, or process it independently or collectively. You can command the device. Software and / or data may be any type of machine, component, physical device, virtual equipment, computer storage medium or device in order to be interpreted by or to provide instructions or data to the processing device. Or may be permanently or temporarily embodied in a signal wave to be transmitted. The software may be distributed over networked computer systems so that they may be stored or executed in a distributed manner. Software and data may be stored on one or more computer readable recording media.

The method according to the embodiment may be embodied in the form of program instructions that can be executed by various computer means and recorded in a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the media may be those specially designed and constructed for the purposes of the embodiments, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks, such as floppy disks. Magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

Although the embodiments have been described by the limited embodiments and the drawings as described above, various modifications and variations are possible to those skilled in the art from the above description. For example, the described techniques may be performed in a different order than the described method, and / or components of the described systems, structures, devices, circuits, etc. may be combined or combined in a different form than the described method, or other components. Or even if replaced or substituted by equivalents, an appropriate result can be achieved.

Claims (14)

Identifying at least one word included in the text; Associating said identification word with identification information including location information associated with said identified word; Identifying relationship information between the identified word and other words included in the text based on the identification information; And Constructing a schema for the text based on the identification information and the relationship information Including, how to build a schema for text. The method of claim 1, And a word contained within the text is itself a character or a contiguous string in the text that has meaning. The method of claim 1, Identifying the at least one word, Determining target words valid for building the schema among words included in the text; And Extracting the determined target words Including, how to build a schema for text. The method of claim 3, The associating the identification information may include associating identification information of each of the extracted target words with a corresponding target word, The identifying of the relationship information may include identifying relationship information between at least two target words among the extracted target words, The step of building the schema, Determining a plurality of nodes associated with each of the extracted target words and each of the identification information associated with the extracted target words; Based on the identified relationship information, connecting a first node associated with a first target word of the extracted target words and a second node associated with a second target word associated with the first target word; And Connecting the node representing the identification information of the first target word with the first node, and connecting the node representing identification information of the second target word with the second node. Including, how to build a schema for text. The method of claim 4, wherein Wherein the first target word is a word included in a search word input by a user. The method of claim 1, And information relating to the association of the identified words with the identification information is stored in a database. The method of claim 1, The identification information includes at least one of Uniform Resource Identifier (URI) information of the identified word, information representing a part of speech of the identified word, and linguistic information of the identified word. How to build a schema for it. The method of claim 7, wherein The identification information is information indicating a position of the text in a document containing the text, information indicating a position in the text of the identified word, information indicating a stemming result of the identified word, the identified word And at least one of information representing a part-of-speech of the information, information representing the entity name of the identified word, and information representing a tag of the entity name of the identified word. The method of claim 1, If the identified word is a word that is also included in other text different from the text, identifying the relationship information identifies relationship information between the identified word and other words included in the other text, The constructing of the schema may include constructing a schema for the text based on the identification information of the identified word and relationship information between the identified word and other words included in the other text. How to build. The method of claim 1, Wherein said relationship information includes information indicative of a dependency relationship between said identified word and said other word. The method of claim 10, And the dependency relationship is at least one of a spelling relationship, a subject-object relationship, and a mathematical relationship between the identified word and the other word. Identifying at least one word included in the text; Associating said identification word with identification information including location information associated with said identified word; Identifying relationship information between the identified word and other words included in the text based on the identification information; Constructing a schema for the text based on the identification information and the relationship information; and Storing the constructed schema in a database Including, how to build a database of schema for text. Identifying at least one word included in the text; Associating said identification word with identification information including location information associated with said identified word; Identifying relationship information between the identified word and other words included in the text based on the identification information; Constructing a schema for the text based on the identification information and the relationship information; Receiving a search word from a user; And If the search word includes the identified word, providing a search result based on the established schema as a search result for the search request of the user by the search word Including, search results providing method. Identify at least one word included in text, associate the identified word with identification information including location information associated with the identified word, and include in the identified word and the text based on the identification information A control unit for identifying relationship information between different words and building a schema for the text based on the identification information and the relationship information; And A storage unit for storing at least one of information about an association between the constructed schema, the identified word, and the identification information A system for processing information included in text, including.

Images