KR101654717B1 - Method for producing structured query based on knowledge database and apparatus for the same - Google Patents

Abstract

 The present invention relates to a method for converting a query composed of natural language into a structured query, and a method for generating a structured query based on a knowledge database according to the present invention is a method for classifying natural query sentences according to a predetermined analysis unit, Determining an attribute of the query and determining a sentence format of the query using the attribute; And a query module configured to determine a type of correct answer to the query statement based on the determined attribute based on a knowledge database that is a query target of the query statement and to generate a query structured in the sentence format with respect to the knowledge database according to the determined correct answer type . According to the present invention, structured queries can be generated by using syntactic and grammatical analysis results of a natural language query to provide more user-friendly information, and it is possible to provide all the data models that are not restricted to specific data models The query is generated based on the knowledge database, so that the correct answer can be obtained from various raw information. Ultimately, because the machine translates into a recognizable query, it can support human-machine communication through natural language.

Description

 TECHNICAL FIELD The present invention relates to a method and apparatus for generating a structured query based on a knowledge database,

 The present invention relates to a method for converting a query composed of natural language into a structured query.

Currently, the technology that is commonly used in search engines is keyword based. If the user inputs a sentence type query as a natural language, the main keywords are extracted through the natural language processing technique and the web documents containing the keyword are returned from the set of the web documents.

In this case, natural language means Korean, English, Japanese, etc. used by ordinary people in daily life, and contrasts with a programming language and data query language understood by machines. Although natural language does not necessarily have a perfect grammar, there is no big problem in human communication. However, in case of a search engine, only the result according to the inclusion of keywords is presented. Therefore, if a user inputs a specific query in order to obtain desired information, And the accuracy of information provided is also reduced.

That is, a system for converting a knowledge database into a recognizable query is needed in order to make the knowledge database know the natural language inputted by the user. In this case, simple conversion according to the determined rules to convert the natural language into the structured query may cause a logical error according to various expression methods of various users. Also, it may be applied according to the structure of the knowledge database to obtain the correct answer to the query An impossible problem arises.

It is an object of the present invention to provide a method for converting a natural language-based query into a structured query that can include all data models rather than a query limited to a specific data model.

More specifically, the present invention aims to provide a method of converting a query into a structured query by using a keyword in a query, as well as syntactic and grammatical analysis results of a query to understand the question in more detail.

 According to an embodiment of the present invention, there is provided a method for generating a structured query based on a knowledge database, the method comprising the steps of: dividing a natural language query according to a predetermined analysis unit to determine attributes of a separated unit syntax; Determining a sentence form of the inquiry; And a query module configured to determine a type of correct answer to the query statement based on the determined attribute based on a knowledge database that is a query target of the query statement and to generate a query structured in the sentence format with respect to the knowledge database according to the determined correct answer type .

Wherein the determining step comprises: dividing the natural language query into morpheme units and associating the morphemes with each other according to a linguistic association between the separated morphemes; and the generating step comprises: It is desirable to determine the type of correct answer.

Wherein the determining step separates the natural language query sentence into a frame having a predetermined semantic unit and analyzes a relationship between the natural language query sentence and a correct answer to the query sentence of the frame, It is preferable to determine the correct answer type for the query.

The determining step may analyze an attribute of an entity name included in the natural language query, and the generating step may determine a correct answer type for the query based on the attribute of the entity name.

The generating step may derive a type candidate of the correct answer to the query statement and determine the type of the correct answer from the type candidate of the derived correct answer with reference to the knowledge database.

The generating step may generate the basis of the structured query using the type of the correct answer and convert the basis of the structured query into the structured query in a predetermined rule according to the determined sentence format .

Preferably, the generating step further specifies the structured query using time or spatial information included in the natural language query.

 According to the present invention, structured queries can be generated by using syntactic and grammatical analysis results of a natural language query to provide more user-friendly information, and it is possible to provide all the data models that are not restricted to specific data models The query is generated based on the knowledge database, so that the correct answer can be obtained from various raw information. Ultimately, because the machine translates into a recognizable query, it can support human-machine communication through natural language.

1 is a diagram illustrating an entire system including a structured query generation apparatus based on a knowledge database according to an embodiment of the present invention.
2 is a flowchart illustrating a method for generating a structured query based on a knowledge database according to an embodiment of the present invention.
FIG. 3 is a flowchart illustrating the determination step of the knowledge database-based structured query generation method according to an embodiment of the present invention in more detail.
FIG. 4 is a flowchart illustrating in greater detail a generation step of a knowledge database based structured query generation method according to an exemplary embodiment of the present invention.

The following merely illustrates the principles of the invention. Therefore, those skilled in the art will be able to devise various apparatuses which, although not explicitly described or shown herein, embody the principles of the invention and are included in the concept and scope of the invention. It is also to be understood that all conditional terms and examples recited in this specification are, in principle, expressly intended for the purpose of enabling the inventive concept to be understood, and are not intended to be limiting as to such specifically recited embodiments and conditions .

BRIEF DESCRIPTION OF THE DRAWINGS The above and other objects, features and advantages of the present invention will become more apparent from the following detailed description taken in conjunction with the accompanying drawings, in which: .

In the following description, a detailed description of known technologies related to the present invention will be omitted when it is determined that the gist of the present invention may be unnecessarily blurred. Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a diagram illustrating an entire system including a structured query generation apparatus based on a knowledge database according to an embodiment of the present invention.

Referring to FIG. 1, a system according to an embodiment of the present invention includes a query processing unit 100 for receiving a query from a user or an external system 10, a query analysis unit 200 for analyzing a query query based on the input natural language, And a query generation unit 300 for generating a structured query through a result, and includes a knowledge database 400 based on generation of a query.

In this embodiment, the query is a query based on a natural language inputted from a user or an external system. The query refers to Korean, English, Japanese, and the like used by ordinary people in everyday life, and contrasts with programming languages and data query languages understood by machines.

Natural language does not necessarily have perfect grammar, but there is no big problem in human communication. Meanwhile, the query based on the knowledge database 400 generated as a result of the system proposed in the present invention means a formal and structured language understood by the machine.

Currently, the most commonly used techniques in search engines are keyword based. When a user inputs a sentence type query, key words are extracted through a natural language processing technique and the web documents including the keyword are returned from the set of the web documents. In addition, the Apple or Samsung interface handles questions and commands specific to data, applications, etc. on the mobile device.

The system of the present invention, on the other hand, transforms the query into a knowledge database (400) query that can encompass all database (400) models, rather than queries limited to a particular database (400) model. In addition, we use the results of syntactic and grammatical analysis of questions as well as keywords to understand and translate the questions more specifically.

In the present invention, the knowledge database 400 to be a query object may be a Semantic Web. The Semantic Web is a relationship between information and resources related to resources such as a web document, various files, various application services, and the like in a distributed environment such as the Internet currently used - ontology information capable of mechanically processing semantics And the like. `

The knowledge database 400 is the most commonly used model for data management, and the database management system (DBMSs) that manage the knowledge database 400 supports a structural query language. However, these data mace management systems are likely to incur errors that result in different results than originally intended through the conversion process because of the different viewpoints and linguistic characteristics of the data.

Therefore, the query analyzer 200 and the query generator 300 according to the present invention perform a method of generating a structured query based on the knowledge database 400, Into a knowledge base query that can cover the model. In addition, we use the results of syntactic and grammatical analysis of questions as well as keywords to understand and translate the questions more specifically.

This will be described in more detail below with reference to Figs. 2 to 4. Fig.

FIG. 2 is a flowchart illustrating a structured query generation method based on the knowledge database 400 according to the present invention.

Referring to FIG. 2, the structured query generation method based on the knowledge database 400 according to the present embodiment includes a determination step S100 and a generation step S200.

In the present embodiment, the determination step S100 determines the attribute of the unit sentence separated by separating the natural language query according to a predetermined analysis unit, and determines the sentence form of the query using the attribute.

In other words, the query analyzer 200 analyzes the question generated in the natural language transmitted from the question processor 100 linguistically.

The determination step S100 by the query analysis unit 200 will be described in more detail with reference to FIG.

The determination step S100 according to the present embodiment includes a morphological analysis step S110, a tree generation step S120 through a syntax analysis, a query statement and a correct answer relation step S130 through a frame analysis, an object name analysis step S140, A sentence format analysis step S150, and an analysis result integration step S160.

In this embodiment, the morpheme analysis step (S110) analyzes the morpheme of the natural language question by the morpheme analysis unit 210 to generate a morpheme list.

Hereinafter, in the present embodiment, when a question statement "What is the world's first organization of women's activists of the socialist and nationalist families in Yeoncheon, Gyeongsangbuk-do in August 1927" Will be described.

Accordingly, the morpheme analysis step (S110) analyzes the natural language query in morpheme units. That is, it can be analyzed as shown in Table 1.

JN: 5: 6] `1927` [`1927` / SN: 7] for the first time [` first `/ NN: 3: 5 + Socialism `/ NN: 14: 15]` `Socialism` `/ NN: 14: 15` `` Nationality `/ NN: 25: 29]` `family` `/` Nationality `/ Nationality` / NN: 25: NN: 37: 40 + `` `` `` `` `` `` `` `` `` `` ` / J: 41: 42] 'World' ['World' / NN: 43: 45] [JNN: 55: 57] "from the Yeongcheon / NN: 55: 57 +" to the "integrated" / N: 60: 62 + `Sun` / X: 62: 63] `Organizations` [`Organization` / NN: 64: 66 + `/ NN: 68: 70 +` `/ J: 70: 71 +`? `/ S: 71: 72]

In this case, NN means a noun, and J is a tag that means an investigation. For example, 'world' [world / NN: 0: 2] Means that it is in the second, and `first 'means it is a combination of nouns and investigations.

In the tree generation step (S120) through parsing, the syntax analysis unit (220) analyzes the syntax of the natural language query to generate a syntax tree. The generated data structure at this time can be expressed as shown in Table 2 below.

{"id": 0, "text": "world", "head": 1, "label": "NP", "mod"
{"id": 1, "text": "first", "head": 14, "label": "NP_AJT", "mod"
{"id": 2, "text": "1927", "head": 3, "label": "NP", "mod"
{"id": 3, "text": "August", "head": 14, "label": "NP_AJT", "mod"
{"id": 4, "text": "socialism", "head": 5, "label": "NP", "mod"
{"id": 5, "text": "series and", "head": 7, "label": "NP_CNJ", "mod"
{"id": 6, "text": "nationalism", "head": 7, "label": "NP", "mod"
{"id": 7, "text": "series", "head": 9, "label": "NP_MOD"
{"id": 8, "text": "female", "head": 9, "label": "NP", "mod"
{"id": 9, "text": "activists", "head": 14, "label": "NP_SBJ"
{"id": 10, "text": "world", "head": 11, "label": "NP", "mod"
{"id": 11, "text": "first", "head": 14, "label": "NP_AJT", "mod"
{"id": 12, "text": "Gyeongsangbuk-do", "head": 13, "label": "NP", "mod"
{"id": 13, "text": "from Yeongcheon", "head": 14, "label": "NP_AJT", "mod"
{"id": 14, "text": "integrate", "head": 15, "label": "VP", "mod": [1, 3, 9, 11,
{"id": 15, "text": "organized", "head": 16, "label": "VP_MOD"
{"id": 16, "text": "group?", "head": -1, "label": "ROOT", "mod"

Table 2 shows the syntactic relationship between the words. Id is the id of the word. Text means a word. Head means parent word, Label means the syntactic role of the word, and Mod means child word. In other words, the most upper - level word of this example query has the word `16 group`, which serves as the ROOT, and the child word `group` has 15 words. The relationship between these connected words is called a syntax tree.

In step S130, the frame analyzing unit 230 separates the natural language query into a frame, which is a predefined semantic unit, to determine the relationship between the query and the correct answer to the query in the frame Analyze.

At this time, the relation between the query and the frame may be expressed as Table 3 according to the above embodiment as a condition for constraining the correct answer.

C: 0 / X: AT_VARIABLE / PT_SUPERATIVE: PT_SUPERATIVE / world first: AT_STRING
C: 1 / X: AT_VARIABLE / PT_TEMPORAL: PT_TEMPORAL / August 1927: AT_STRING
C: 2 / X: AT_VARIABLE / PT_SUPERATIVE: PT_SUPERATIVE / world first: AT_STRING
C: 3 / X: AT_VARIABLE / PT_LOCATION: PT_LOCATION / from Yeongcheon: AT_STRING
C: 4 / X: AT_VARIABLE / PT_LOCATION: PT_LOCATION / Gyeongsangbuk-do: AT_STRING
C: 5 / X: AT_VARIABLE / PT_TYPE: PT_TYPE / Organization: AT_STRING
C: 6 / Athletes: AT_STRING / Integrated: PT_LEXICAL / X: AT_VARIABLE

Table 3 shows the result of analyzing the frame of the query sentence, and C denotes the constraint id of the variable (X) appearing in the frame. In the case of the 5th condition, for example, C: 5 means that the type (PT_TYPE) of the variable (AT_VARIABLE) X that is the correct answer by the query is `group`. The sixth condition means `the activists` integrated `X '. Also, condition 1 means that it happened in August 1927.

In the object name analysis step S140, the object name analysis unit 240 analyzes the object name attribute included in the natural language query. At this time, the entity name can be identified through the dictionary managing the entity name, and the attribute of the identified entity name is analyzed.

The result of analyzing the object name according to the above query can be expressed as shown in Table 4.

August 1927: DT, Socialism: TR, Nationalism: TR, Gyeongsangbuk-do: LC, Yeongcheon: LC

The name of the object that represents the time (DT) named `August 1927` as the recognized object name. Analyzes its properties with the object name representing the space (LC) called `Gyeongbukbukdo`.

In the sentence form analysis step S150, the sentence form analyzer 250 grasps the subject predicate relation of the sentence through the attribute of the morpheme and the attribute of the entity name, and analyzes the sentence form according to a predetermined format criterion. In this embodiment, the structure of the structured query in which the analyzed sentence format is generated can be determined.

In the following analysis result integration step S160, the analysis result integration unit S260 includes the morphological analysis unit 210, the syntax analysis unit 220, the frame analysis unit 230, the object name analysis unit 240, (250) and transmits the combined result to the query generation unit (300).

Hereinafter, the generation step S200 will be described in more detail with reference to FIG.

Referring to FIG. 4, the generating step S200 includes a correct answer type candidate obtaining step S210, a correct answer type determining step S220, a query basic creating step S230, a structured query creating step S240, A query specification step S250, and a query integration step S260.

In the correct answer type candidate derivation step S210, the correct answer type determination unit 310 determines the type of the correct answer to obtain from the natural language question based on the language analysis result. First, a number of correct answer candidates are derived through the language analysis results.

In the present embodiment, information on the condition of the variable can be obtained from the result of the frame analyzer 230. In the present embodiment, `C: 5 / X: AT_VARIABLE / PT_TYPE: PT_TYPE / You can get information that the type is related to the `group 'character. Several types of these types can be extracted from the frame analysis results, and this can be determined as a candidate for the correct answer type.

Next, in the correct answer type determination step S220, the correct answer type determination unit 310 refers to the knowledge database 400 to determine the correct answer type. That is, the data structures of the knowledge database 400 to be subjected to the query can be referred to, and the most suitable correct answer type can be selected. Accordingly, in this embodiment, since the correct answer type corresponding to the structural characteristic of each knowledge database 400 is found and a structured query is generated therefrom, a knowledge database (not shown) 400), so that correct answers can be obtained from a variety of raw information

In the present embodiment, candidates including a plurality of correct answer type candidates derived above are searched in the knowledge database 400. If there is one candidate, select it. If not, select the best candidate. For example, if two or more occur, candidates with attributes related to the PT_LEXICAL value of the frame analysis result can be selected.

The correct answer type basic structured query basic creation step S230 generates a basic of a knowledge base based structured query based on the correct answer type derived by the correct answer type determination unit 310. [

In the structured query generation step S240, the frame conversion unit 320 generates a knowledge base query with a predetermined rule according to the sentence format of the natural language query as a result of the query analysis.

The structured query generated according to the above query can be expressed as shown in Table 5 below.

? stmt rdf: type rdf: Statement.
? stmt rdf: object? X.
? stmt rdf: predicate xbp: founderOf.
? stmt rdf: subject? ins_1393423910.
? ins_1393423910 xbp: name? ins_name_1393423910.
FILTER regex (? Ins_name_1393423910, "activist")

In the query specification step S250, the additional information conversion unit 330 specifies the structured query using time or spatial information included in the natural language query.

The materialized structured query generated according to the above query can be expressed as shown in Table 6 below.

? stmt xbp: occursOn? tem_966484835.
? tem_966484835 xbp: name? tem_name_966484835.
FILTER regex (? Tem_name_966484835, "August 1927")
? stmt xbp: occursIn? loc_144315937.
• loc_144315937 xbp: name • loc_name_144315937.
FILTER regex (? Loc_name_144315937, "Yeongcheon")
? stmt xbp: ranking? sup_1008915696.
FILTER regex (? Sup_1008915696, "world first")

In the query integration step S260, the query combination unit 340 determines a part of the knowledge database 400 based query generated in the correct answer type determination unit 310, the frame conversion unit 320, and the additional information conversion unit 330, And transmits the generated query to the query processing unit 100.

The combined structured queries generated according to the above query can be expressed as shown in Table 7 below.

SELECT? Name
WHERE {
? X xbp: name? Name.

? X rdf: type xbc: Organization.

? stmt rdf: type rdf: Statement.
? stmt rdf: object? X.
? stmt rdf: predicate xbp: founderOf.

? stmt rdf: subject? ins_1393423910.
? ins_1393423910 xbp: name? ins_name_1393423910.
FILTER regex (? Ins_name_1393423910, "activist")

? stmt xbp: occursOn? tem_966484835.
? tem_966484835 xbp: name? tem_name_966484835.
FILTER regex (? Tem_name_966484835, "August 1927")
? stmt xbp: occursIn? loc_144315937.
• loc_144315937 xbp: name • loc_name_144315937.
FILTER regex (? Loc_name_144315937, "Yeongcheon")
? stmt xbp: ranking? sup_1008915696.
FILTER regex (? Sup_1008915696, "world first")
}

The structured query based on the knowledge database 400 generated through the query generation unit 300 is transmitted again to the query processing unit 100 and finally returned to the external system or user. When the external system or the user requests the generated query to the knowledge database 400, the knowledge database 400 returns the correct answer such as " anniversary. &Quot;

According to the present invention, structured queries can be generated by using syntactic and grammatical analysis results of a natural language query to provide more user-friendly information, and it is possible to provide all the data models that are not restricted to specific data models The query is generated based on the knowledge database, so that the correct answer can be obtained from various raw information. Ultimately, because the machine translates into a recognizable query, it can support human-machine communication through natural language.

It will be apparent to those skilled in the art that various modifications, substitutions and substitutions are possible, without departing from the scope and spirit of the invention as disclosed in the accompanying claims. will be.

Therefore, the embodiments disclosed in the present invention and the accompanying drawings are intended to illustrate and not to limit the technical spirit of the present invention, and the scope of the technical idea of the present invention is not limited by these embodiments and the accompanying drawings . The scope of protection of the present invention should be construed according to the following claims, and all technical ideas within the scope of equivalents should be construed as falling within the scope of the present invention.

Claims (8)

Determining a property of the separated unit syntax by separating the natural language query statement according to a predetermined analysis unit, and determining the sentence form of the inquiry query using the attribute; And
Determining a type of a correct answer to the query according to the determined attribute based on a knowledge database to be a query object of the query,
And generating a structured query in the sentence format for the knowledge database according to the determined correct answer type,
Wherein the natural language query sentence is divided into morpheme units and a syntactic tree is generated, and the natural language query sentence is divided into frames, each of which is a predefined semantic unit, using the syntax tree to restrict the correct answer of the natural language query The condition is derived,
Wherein the generating step generates a structured query according to a selected correct answer type with reference to a data structure of the knowledge database according to a condition limiting the correct answer. delete delete The method according to claim 1,
Wherein the determining includes analyzing attributes of entity names included in the natural language query,
Wherein the generating step determines the correct answer type for the query statement according to the attribute of the entity name. The method according to claim 1,
Wherein the generating step derives a type candidate of a correct answer to the query statement,
And the type of the correct answer is determined from the type candidate of the correct answer derived by referring to the knowledge database. The method according to claim 1,
Wherein the generating step generates the basis of the structured query using the type of the correct answer,
And converting the basis of the structured query into the structured query in a predetermined rule according to the determined sentence format. The method according to claim 1,
Wherein the step of generating the structured query specifies the structured query using time or spatial information included in the natural language query. A query analyzer for determining the attribute of the separated unit syntax by separating the natural query sentence according to a predetermined analysis unit and determining the sentence form of the query sentence using the attribute; And
Determining a type of a correct answer to the query according to the determined attribute based on a knowledge database to be a query object of the query,
And a query generation unit configured to generate a structured query in the sentence format for the knowledge database according to the determined correct answer type,
Wherein the query analyzing unit generates a syntax tree by separating the natural language query sentence into morpheme units and separates the natural language query sentence into frames of a predefined semantic unit using the syntax tree to restrict the correct answer of the natural language query ≪ / RTI >
Wherein the query generator generates a structured query based on the selected correct answer type with reference to the data structure of the knowledge database according to a condition constraining the correct answer.

Images