KR101593214B1 - Method and system for searching by using natural language query - Google Patents

Abstract

The present invention relates to a method for searching a natural language query of a logical and physical binary structure, wherein a semantic relation between morphemes constituting the natural language query is analyzed with respect to a natural language query, Up search language translation step; If the attribute structure of the mesh-up search language is a logical / physical duplex structure requiring result data through a logical operation using physical data, the mesh-up search language may be a dual structure including a predetermined data search rule information A target search language conversion step of converting a target search language into a target search language; And a retrieval step of performing data retrieval in a predetermined knowledge database based on the data retrieval rule information included in the target retrieval language of the double structure.

Description

TECHNICAL FIELD [0001] The present invention relates to a method and system for searching natural language queries using a logical physical double structure,

More specifically, the present invention proposes a search language capable of actively responding to various search requests of semantic-tagged natural language queries and provides flexible information through a logical / physical structure double structure of search languages And more particularly, to a method and system for searching natural language queries.

Ontology defines a semantic relationship between real-world objects and provides semantic specification of information for the computer to access information derived from complex relationships between objects. It is used for artificial intelligence, information retrieval, ubiquitous, e-commerce, etc. It can be applied in various prefectures.

In particular, the ontology expresses a variety of information on the web in a semantic relationship between objects, thereby obtaining a logical query result that can not be found in an existing keyword-based search in the information search field. For example, when using an ontology, you can get the results of a query such as "Where is the population of more than 70 million, and where is the capital of the neighboring country?"

Such ontology information retrieval is more useful in three aspects than keyword based retrieval. First, the keyword-based search results provide only the documents containing the input keywords as search results. However, the ontology information search can directly provide the specific information that the user wants to search and attribute information thereof. Second, keyword based search can not guarantee the search result reflecting the relation between input keywords. However, the ontology information search can find information from the relation with other objects that can characterize the search target. Third, keyword - based search can search only existing information, but searching ontology information can deduce new information that did not exist in the past by searching for a new relation from the relationship between objects.

However, this ontology information retrieval is a great help in user's convenience by directly searching for a query object desired by the user, but it is difficult to access the ontology query language because it is not familiar to the general user. As a result, the ontology becomes ineffective because the user has to learn the structure of the ontology and the language of the ontology to access the information constructed by the ontology.

On the other hand, natural language is enough to express the logical meaning expressed in ontology query language, and is familiar to the general user. Therefore, if the ontology information can be accessed by the natural language query, it is possible to improve the efficiency of the ontology information search by making the access of the general user easy.

In order to access ontology information by natural language query, it is required to convert natural language query into ontology query language. To do this, first, structure and meaning of natural language query are analyzed by using natural language processing technology, We need to find the object and relation of the ontology corresponding to the natural language expression and convert the natural language query into the ontology query language according to the grammar structure of the ontology query language.

However, this method has a problem that an ontology query language which can not be accessed due to errors such as structural analysis and semantic analysis on a natural language query occurs. More specifically, there is no guarantee that general natural language processing techniques have all the language resources needed to process natural language expressions that can access the ontology to be searched, and it can lead to false analysis results due to minor grammatical errors. An ontology query term that can not be generated may occur.

As a result, according to the existing retrieval method, the user has to learn the types of queries that the system can process while experiencing many trial and error about an error query. As a result, the user who uses the ontology- It does not know what query is useful because it does not know the type of query that the processing technology can accept and the specific content of the information constructed in the ontology.

Also, from the service provider 's point of view, it means that it can not provide all of the constructed ontology information.

However, most users want to intuitively use the information on the ontology prepared by the service provider rather than knowing the information about the query and the method of inputting the query. As a result, , The utilization and satisfaction of the system is reduced.

Therefore, research on Natural Language Understanding (NLU) is continuing to improve the utility and accuracy of information retrieval. Natural language technology is intended to grasp the user's intention of the user on the 'user question'. It is the input of the question which is uttered by voice or inputted as text, and the extracted list by understanding the intention of the question in the natural language technique, .

Meaning Using the tagged natural language, it is possible to inquire that the complexity is varied and the user's intention can be revealed more clearly. However, there is a difficulty in building a knowledge database of a data structure that can reflect not only a technique for accurately grasping user intention contained in a query word but also various needs of a user, based on a natural language query.

An object of the present invention is to solve the above-mentioned problems, and it is an object of the present invention to define a search domain and an attribute structure by morpheme analysis constituting a query word for a complex natural language query having a logical and physical double structure, To a search language corresponding to the physical structure of the database.

It is another object of the present invention to minimize the database capacity of a search system and the load of a search engine by providing search results matching user intent using physical data stored in an existing knowledge database.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory and are not restrictive of the invention, unless further departing from the spirit and scope of the invention as defined by the appended claims. It will be possible.

According to an aspect of the present invention, there is provided a method of searching for a natural query according to an embodiment of the present invention, comprising: converting a query term into a mesh-up search language through analysis of semantic relations between morphemes constituting the natural- A mesh-up search language conversion step; If the attribute structure of the mesh-up search language is a logical / physical duplex structure requiring result data through a logical operation using physical data, the mesh-up search language may be a dual structure including a predetermined data search rule information A target search language conversion step of converting a target search language into a target search language; And a retrieval step of performing data retrieval in a predetermined knowledge database based on the data retrieval rule information included in the target retrieval language of the double structure.

The data retrieval rule information according to an exemplary embodiment of the present invention includes attribute information of one or more physical data required to derive result data for a mesh-up retrieval language which is a logical / physical double structure, And may include rule information.

The retrieving step according to an embodiment of the present invention includes: deriving one or more physical data from a knowledge database based on the data retrieval rule information included in the target retrieval language; And deriving result data according to the calculation rule information using the derived one or more physical data.

According to another aspect of the present invention, there is provided a natural language query query system for analyzing a natural language query through a semantic relation analysis between morphemes constituting a natural language query input through a user device, A natural language processing engine for converting the natural language; A search engine for converting the mesh-up search language generated by the natural language processing engine into a target search language, which is a database search language, and providing search and result data based on the target search language; And a knowledge database for storing knowledge information and language conversion information, wherein the search engine is a logical / physical dual structure in which the attribute structure of the mesh-up search language requests result data through logical operation using physical data , It converts the mesh-up search language into a target search language of a double structure including predetermined data search rule information, and performs search based on the data search rule information.

The search engine according to an embodiment of the present invention includes a target search language conversion module for analyzing a domain and an attribute structure of a mesh-up search language generated by the natural language processing engine and converting the domain and the attribute structure into a target search language based on an attribute structure; And a data derivation module for retrieving and deriving physical data from the knowledge database based on the target search language generated by the target search language conversion module and generating result data.

In the case where the attribute structure of the mesh search language is a logical / physical duplex structure, the target search language conversion module according to an embodiment of the present invention may further include: You can convert to the target search language of the structure.

At this time, the data derivation module derives one or more physical data from a knowledge database based on the data retrieval rule information included in the target retrieval language of the double structure converted by the target retrieval language conversion module, It is possible to derive result data according to the calculation rule information using one or more physical data.

It is to be understood that both the foregoing general description and the following detailed description of the present invention are exemplary and explanatory and are intended to provide further explanation of the present invention by those skilled in the art. And can be understood and understood.

According to an embodiment of the present invention, a flexible query structure processing method is proposed, and a high quality search result can be provided based on natural language query characteristics.

According to an embodiment of the present invention, a database capacity of a search system and a load of a search engine are minimized by providing search results matching user intent using physical data stored in an existing knowledge database.

BRIEF DESCRIPTION OF THE DRAWINGS The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of the specification, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention.
1 is a diagram illustrating an example of a logical / physical dual structure based search system according to an embodiment of the present invention.
2 is a diagram illustrating an example of an NQL generated based on a user query according to an embodiment of the present invention.
3 is a diagram illustrating an example of a process of providing a search result based on a mesh-up search language in a search engine according to an embodiment of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS The present invention is capable of various modifications and various embodiments, and particular embodiments are illustrated in the drawings and described in the following description of the invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the accompanying drawings. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The following detailed description, together with the accompanying drawings, is intended to illustrate exemplary embodiments of the invention and is not intended to represent the only embodiments in which the invention may be practiced. The following detailed description includes specific details in order to provide a thorough understanding of the present invention. However, those skilled in the art will appreciate that the present invention may be practiced without these specific details.

The present invention proposes a search language capable of actively responding to various search requests of semantic-tagged natural language queries, and to a natural language query searching method and system for providing flexible information through a logical / physical structure binary structure of a search language.

1 is a diagram illustrating an example of a natural language query-based retrieval system according to an embodiment of the present invention.

Referring to FIG. 1, a search system 100 according to an embodiment of the present invention includes a natural language processing engine (NLU) based on a Natural Language Understanding (NLU) for analyzing a user query input through a user device 110 120, a search engine 130, and a knowledge database 140.

The natural language processing engine 120 analyzes a complex query term based on the semantically tagged natural language inputted through the user device 110 and converts it into a Naver Contents Repository-mashup Query Language (NQL) through semantic analysis, The search engine 130 derives the query-requested information from the pre-built knowledge database based on the NQL generated by the natural language processing engine 120 and provides it to the user device 110.

Hereinafter, each configuration of the search system according to the embodiment of the present invention will be briefly described.

The natural language processing engine 120 includes a semantic analysis module 121 for performing semantic analysis of a user query and an NQL conversion module 122 for converting a natural language-based user query into an inferable NQL according to a semantic analysis result.

The semantic analysis module 121 analyzes the user query word, classifies it into morpheme units, and analyzes the logical / physical double structure of the user query word by grasping the semantic relation between morphemes.

For example, if the user query term is "current age of a specific person A ?, ", the corresponding query term is a query phrase for requesting physical data" And the result of the query that requests the physical data of "Year?" Can be analyzed. In other words, through analyzing the semantic relationship between the morphemes of the query words, the "current age" can be understood as a logical / physical dual structure including the physical data such as "date of birth" and "current year" internally.

The NQL conversion module 122 is a module for converting a user query term based on a salmon into a search language NQL and can infer one or more query phrases included in the user query based on the analysis result performed by the semantic analysis module 121 NQL of the tree structure is created.

In this specification, NQL is a search language for specifying a service unit attribute by analyzing a query word, and can be defined as a language designed to query a knowledge database by meaning tagging a user query expressed in a natural language. NQL consists of a dual structure that externally represents logical attributes and internally generates search results using the physical attributes of the knowledge database. In other words, NQL is a language for building a unique logical structure that does not conform to the physical structure of the knowledge database, and defines the relationship between the logical structure and the physical structure through the setting so that appropriate data can be provided for each type of query and service type. For example, if the user query includes the phrase "how old are you?", The converted NQL can be specified as "age".

The NQL of the tree structure generated by the NQL conversion module 122 is as illustrated in FIG. This will be described later with reference to Fig.

Next, the search engine 130 performs information search from the knowledge database 140 based on the natural language based NQL generated by the natural language processing engine 120, and is based on the mesh-up technology.

Specifically, the search engine 130 includes an OQL conversion module 131 for converting the NQL sent from the natural language processing engine 120 into a target query language (OQL), a knowledge database 140, A data derivation module 132 for retrieving and deriving data corresponding to the retrieved data, and a data processing module 133 for processing and providing the retrieved retrieved data in accordance with the user's intention.

In this specification, OQL can be defined as a language optimized for searching a knowledge database, and OQL uses a physical structure used by a knowledge database. For example, in the above embodiment, an OQL expressed as " a search language for calculating an age using a current year value and a birth year value " for a property language NQL "age " . ≪ / RTI >

The knowledge database 140 includes an OQL DB 141, a knowledge information DB 142, and a data processing rule DB 143 for storing attribute information and rule information required for OQL conversion. At this time, it is preferable that the OQL DB 141 stores the property definition rules necessary for solving the logical / physical dual structure of NQL so as to express the logical / physical dual structure of NQL at the time of OQL conversion.

The OQL conversion module 131 converts the NQL provided from the natural language processing engine into the OQL based on the OQL DB 141. The OQL conversion module 131 can derive desired data as physical attribute values of the knowledge database based on the logical domain and the attributes of the NQL OQL.

In one embodiment, when the NQL attribute for a specific person to be queried is in accordance with a physical structure such as "date of birth", "year of debut", the OQL transformation module 131 determines the physical structure Based OQL.

On the other hand, if the NQL attribute for a specific person to be queried is based on a logical structure that must derive a result based on two or more pieces of information such as "current age", "de age age", " Can be converted into an OQL based on a logical / physical structure based on an attribute definition rule pre-stored in the OQL DB 141. [

For example, if the "current age" of the person is requested, NQL specifies "age" and OQL according to the logical / physical structure specifies the physical data "birth year" and "current year" Quot; current age "can be converted to be expressed in a language for calculating the" current age " As another example, in the case of NQL "DeveNaiNai ", the OQL can be converted into a language that computes the" DeVeNiNa "using the physical data of the person" date of birth "

Accordingly, the data derivation module 132 searches the knowledge information DB 142 based on the converted OQL from the OQL transformation module 131 to derive physical data. In addition, the data derivation module 132 may derive data consistent with the logical structure using the derived one or more physical data.

According to the embodiment described above, when the OQL is defined as a language for deriving the "current age" based on the physical data of the "birth year" and the "current year" of a specific person, From the "birth year" and the "current year" data derived from the physical data of the person "birth year" and "current year" from the data file 142 and derives the requested data . That is, even if physical data for the "current age" or "debut year" of the person is not separately stored in the knowledge information DB 142, the query result is obtained using the existing data necessary for deriving the result value to the requested question .

The data processing module 133 processes the OQL data derived from the data derivation module 132 based on the data processing rule DB 143 into an NQL result. Preferably, the data processing module 133 may process the data type provided to the user based on the query type, the multi-level query structure, and the setting according to the query attribute according to the predetermined NQL config rule.

Accordingly, the search engine 130 according to the embodiment of the present invention searches the knowledge DB 142 for data of a logical / physical dual structure in which data provided according to a question or a question type of a user is changed By using the stored physical data, it is possible to minimize the load of periodically updating the data stored in the knowledge information DB 142 and the new data.

2 is a diagram illustrating an example of a mesh-up search language generated based on a user query according to an embodiment of the present invention.

Referring to FIG. 2, the user query language is "What is the movie in which the member of the Girls Generation appeared? &Quot;, and the natural language processing engine normalizes the query word by the morpheme, &Quot;, and the semantic relation between morphemes such as "Girls' Generation < - > Member < Movie" is analyzed and can be expressed as NQL of a hierarchical structure as shown in FIG.

The illustrated hierarchical structure can consist of "person / member" 202 and "person / movie_performed" 203 as "object / name": Girls' Generation "201 and attribute information therefor. The attribute of NQL shown in Fig. 2 can be regarded as a hierarchical structure of OPP (Object / Property / Property) type.

According to an embodiment of the present invention, the mesh-up technique includes an OP (Object / Property), a COP (Condition / Object / Property), and an OPP (Object / Property / Property) for each attribute of a morpheme constituting a query word, Such as OOP (Object / Object / Property), OO (Object / Object), CCO (Condition / Condition / Object) and CCOP (Condition / Condition / Object / Property).

3 is a diagram illustrating an example of a process of providing a search result based on a mesh-up search language in a search engine according to an embodiment of the present invention.

Referring to FIG. 3, when a search engine according to an embodiment of the present invention receives an NQL based on a user query term from a natural language processing engine (S301), it performs an analysis on an attribute and a domain of the NQL (S302).

In the analysis process (S302), it is determined whether the attribute structure of the NQL is a single structure requiring physical data, or whether it is a logical / physical dual structure that requires physical data internally or externally requires a logical result.

As a result of the NQL attribute analysis, if the corresponding NQL is a logical / physical double structure, the OQL is converted into an OQL defining one or more physical data required to derive the logical data (S303, S304).

According to the above-described embodiment, when the NQL property structure is a physical unity structure such as "date of birth "," date of debut ", it can be converted into OQL designating "date of birth" Alternatively, when the NQL property structure is a logical / physical double structure such as "current age ", the" debian age "is computed using the physical data" date of birth " You can convert to OQL, which is defined as doing.

In another embodiment, if the NQL attribute structure is a dual structure requiring data for one or more domains (e.g., movies, broadcasts, performances, etc.) called " Quot ;, " feature film ", "feature film ", " feature broadcast "," performance performance ", and the like.

Furthermore, when the NQL property structure is a dual structure requiring a work in which a specific character appears as a "main actor", a plurality of NQL "major actors", "feature films", " The OQL that is defined to selectively derive only the works whose title is set as the "master" in the "starring work" data obtained by using the physical data. In this case, And a hierarchical structure for searching for a hierarchical relationship as a sub-concept.

As described above, one or more physical data required from the knowledge database is retrieved and derived for the converted OQL according to the logical / physical dual structure attribute of NQL, and the resulting data for NQL is derived using one or more derived physical data (S305).

On the other hand, if the NQL attribute structure is a single structure for requesting physical data in the NQL attribute analysis step S302, the OQL for the physical data referred to by the NQL is converted (S303, S306).

Then, the physical data is derived from the knowledge database based on the OQL converted from the NQL of the single structure (S307).

Finally, the search engine processes the data based on the data derived based on the single structure or the logical / physical double structure to match the user query type and provides the data (S308).

The foregoing description is merely illustrative of the technical idea of the present invention, and various changes and modifications may be made by those skilled in the art without departing from the essential characteristics of the present invention. Therefore, the embodiments of the present invention are not intended to limit the scope of the present invention but to limit the scope of the present invention. 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 thereof should be construed as being included in the scope of the present invention.

Claims (13)

1. A method for searching natural language queries of a logical and physical structure by a natural language query language query system,
A mesh-up search language conversion step of analyzing a semantic relationship between morphemes constituting the natural language query word with respect to a natural language query word and converting the semantic relation into a mesh-up search language based on a predetermined search language conversion rule;
If the attribute structure of the mesh-up search language is a logical / physical duplex structure requiring result data through a logical operation using physical data, the mesh-up search language may be a dual structure including a predetermined data search rule information A target search language conversion step of converting a target search language into a target search language; And
And retrieving data from a predetermined knowledge database based on the data retrieval rule information included in the target retrieval language of the double structure. The method according to claim 1,
Wherein the target search language conversion step comprises:
Analyzing a domain and an attribute structure of the mesh-up search language; And
And determining whether the mesh-up search language is the logical / physical duplex structure according to a result of the analysis. 3. The method according to claim 1 or 2,
The data retrieval rule information includes:
A logical / physical double structure natural language query including at least one attribute information of at least one physical data required to derive result data for a mesh-up retrieval language, which is a logical / physical double structure, and arithmetic rule information using the at least one physical data How to search. The method of claim 3,
The retrieving step comprises:
Deriving one or more physical data from a knowledge database based on the data retrieval rule information contained in the target retrieval language; And
And deriving result data according to the calculation rule information using the derived one or more physical data. The method according to claim 1,
Wherein the target search language conversion step comprises:
Further comprising analyzing a domain and an attribute structure of the mesh-up search language. 6. The method according to claim 1 or 5,
Wherein the target search language conversion step comprises:
Further comprising the step of, when the attribute structure of the mesh-up search language is a single structure requiring physical data, converting the object structure into a target search language of a single structure corresponding to the mesh-up search language. How to search. The method according to claim 6,
The retrieving step comprises:
And retrieving and deriving physical data corresponding to the target search language of the single structure from the knowledge database. A natural language query query system comprising:
A natural language processing engine for analyzing a semantic relation between morphemes constituting a natural language query input through a user device and converting the natural language query into a mesh-up search language; And
A search engine for converting the mesh-up search language generated by the natural language processing engine into a target search language that is a database search language and providing search and result data based on the target search language; And
A knowledge database for storing knowledge information and language conversion information,
The search engine comprises:
If the attribute structure of the mesh-up search language is a logical / physical duplex structure requiring result data through a logical operation using physical data, the mesh-up search language may be a dual structure including a predetermined data search rule information And a search is performed based on the data retrieval rule information. 9. The method of claim 8,
The data retrieval rule information includes:
A logical / physical double structure natural language query including at least one attribute information of at least one physical data required to derive result data for a mesh-up retrieval language, which is a logical / physical double structure, and arithmetic rule information using the at least one physical data Search system. 10. The method according to claim 8 or 9,
The search engine comprises:
A target search language conversion module for analyzing a domain and an attribute structure of a mesh-up search language generated by the natural language processing engine and converting the domain and the attribute structure into a target search language based on an attribute structure; And
And a data derivation module for retrieving and deriving physical data from the knowledge database based on a target search language generated by the target search language conversion module and generating result data. 11. The method of claim 10,
Wherein the target search language conversion module comprises:
If the attribute structure of the mesh-up search language is a logical / physical duplex structure, converting the mesh-up search language into a target search language of a double structure including predetermined data search rule information; Natural language query retrieval system. 11. The method of claim 10,
Wherein the data derivation module comprises:
Deriving one or more physical data from a knowledge database based on the data retrieval rule information included in the target retrieval language of the double structure converted by the target retrieval language conversion module, A natural language query system of a logical physical and binary structure deriving result data according to calculation rule information. 9. The method of claim 8,
The search engine comprises:
A semantic analysis module for analyzing a plurality of morphemes constituting the natural language query and analyzing semantic relations between the morphemes; And
And a mesh-up search language conversion module for generating the mesh-up search language based on the analysis result of the semantic analysis module.

Images