KR101667918B1

KR101667918B1 - Methodand device of providing query-adaptive smart search service

Info

Publication number: KR101667918B1
Application number: KR1020150117003A
Authority: KR
Inventors: 배원식; 정형일; 김태일; 은종진; 김도연
Original assignee: 네이버 주식회사
Priority date: 2015-08-19
Filing date: 2015-08-19
Publication date: 2016-10-21

Abstract

The present invention relates to a method of providing a smart search service and a search service apparatus for implementing the smart search service. The smart search service providing method includes: recognizing a query through natural language processing of a query input by a user; arranging contents of a document related to the query as a corresponding pattern; Discloses a method for providing a search service that exposes snippet information including an overall content so that specific contents required by a user can be easily grasped on the document even with only snippet information.

Description

TECHNICAL FIELD [0001] The present invention relates to a smart search service providing method and a search service apparatus for implementing the same,

The present invention relates to a method of providing a query response type smart search service and a search service apparatus for implementing the method. More particularly, the present invention relates to a method and apparatus for recognizing a query through natural language processing on a query input by a user, And a method of providing a search service that can easily grasp a specific content required by a user on the document by using only snippet information by exposing snippet information including the overall contents of the query.

With the development of ICT technology, modern people are searching for an appropriate answer to questions or business matters by searching through internet search service without limitation of time and space.

FIG. 1 shows an example of a search result of a search service according to the related art. When a user inputs a query content into a search window, the search service server searches the document based on the query content and provides search results according to the search result .

In the case of the conventional search service, as shown in FIG. 1 (a), a general search engine extracts keywords included in the query content input through the search window 11, searches for a document including keywords, A search result 13 that exposes a snippet such as a part of a sentence including the keyword 15 is provided.

A snippet is a simple representation of a web page that a search engine displays to a user. In addition, a link or page title, a short page description, or a thumbnail that renders a preview, along with a snippet, It is also provided.

However, as a result of such a search, only the information level indicating that the keyword is included in the document is searched, and the user accesses the document through the click of the exposed area as a search result, The information provided as search results is poor to find the answers you want with just the results.

In order to provide a more effective search result, additional information 25 (e.g., popularity, reliability, etc.) of the document retrieved in association with the query content input through the search window 21 as shown in FIG. 1 In the retrieval result 23. However, even in the case of such a retrieval service, even among the contents of the retrieved document, a sentence including the keyword 27 included in the query content or a partial content of the paragraph Is only exposed to the search result and only the level of the additional information 25 is such that the number of documents to be accessed can be reduced in order to obtain a solution to be searched by the user. There is still a problem in that the desired information can be obtained by clicking on the exposed area to access the document.

Patent Publication No. 10-2013-0128697

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems of the prior art, and it is an object of the present invention to provide a method and a system for providing information, So that the user can solve the problem that the information provided is insufficient to search for the desired solution only by the search result.

According to an aspect of the present invention, there is provided a method of providing a search service, the method comprising: receiving a query of a user; A document retrieval step of retrieving a document based on the query; A snippet extracting step of extracting a main content associated with the query from at least one of the retrieved documents, extracting a main content associated with the query, and extracting a snippet; And a search result providing step of providing a search result for the query including the extracted snippet.

The method further includes a query type analysis step of analyzing a query type of the query to determine a query domain, wherein the snippet extraction step extracts a snippet pattern in which a content development structure corresponding to the query domain is defined, The snippet pattern may be applied to the document, the main content may be extracted from the document according to the snippet pattern, and the snippet may be extracted by collecting the main content.

Further, the snippet extracting step may include a snippet candidate extracting step of extracting a snippet candidate by extracting main contents from the document according to the snippet pattern by applying the snippet pattern to the document and extracting the snippet candidate; And a snippet selection step of selecting a snippet from the snippet candidate through morphological analysis and semantic analysis of the snippet candidate.

Preferably, the query type analysis step includes a query determination step of extracting an element from the query, and determining a clue element and a keyword element corresponding to the clue element in the element through a relation dictionary; A query type determination step of determining a query type according to the combination of the clue element and the keyword element; And a domain determination step of determining a query domain corresponding to the query type.

And a snippet pattern DB constructing step of constructing a snippet pattern DB storing a domain-specific snippet pattern, wherein the snippet candidate extracting step extracts a snippet pattern corresponding to the query domain from the snippet pattern DB .

Wherein the step of constructing the snippet pattern DB includes a document collection step of collecting a document and analyzing the document by each element by performing a natural language process on the collected document; Determining a clue element and a keyword element corresponding to the clue element in the element through a previously held relation dictionary, and determining a snippet domain according to the combination of the clue element and the keyword element; A document analysis step of analyzing a structure of an area including the element; A snippet pattern generation step of determining an association between the elements and generating a snippet pattern by combining the analyzed structures based on the association; And a snippet pattern storing step of storing the snippet pattern in the snippet pattern DB in association with the snippet domain.

Further, the extracting of the snippet candidates may include searching a document based on the elements included in the query, performing natural language processing on the retrieved document, and analyzing the extracted elements by respective elements. And extracting the snippet candidate by applying the extracted snippet pattern to the analyzed document, extracting the content corresponding to the query from the document according to the snippet pattern, collecting the collected content, and extracting the snippet candidate.

The step of extracting a snippet candidate may further include extracting a plurality of snippet patterns corresponding to the query domain, extracting a plurality of snippet candidates by applying each snippet pattern to one document, And may further include a snippet candidate processing step of collecting snippet candidates and integrating them into one snippet candidate.

In addition, the snippet pattern generation step may be performed by separating domain specific snippet patterns corresponding to a specific domain and common snippet patterns corresponding to all domains.

Further, the extracting of the snippet candidates may include collecting documents related to the query, analyzing the documents by respective elements by performing a natural language process on the collected documents, Determining a clue element corresponding to the clue element and a keyword element corresponding to the clue element in the element through a previously held relation dictionary, determining a domain according to the combination of the clue element and the keyword element, and comparing the domain with the query domain; A document analysis step of analyzing a structure of an area including the element; A snippet pattern generation step of determining an association between the elements and generating a snippet pattern by combining the analyzed structures based on the association; And extracting snippet candidates by applying the extracted snippet pattern to the analyzed document, extracting contents corresponding to the query in the document according to the snippet pattern, collecting the contents, and extracting the snippet candidates.

Preferably, the snippet extracting step may include calculating a weight by a morphological analysis of the snippet candidate, calculating relevance by semantic analysis of the snippet candidate, and extracting a snippet candidate based on the weight and the degree of relevance, Can be filtered to extract the snippet.

Here, the snippet extracting step may include a step of analyzing the shape of the snippet candidate, the number of times of application of the snippet pattern applied to the extraction of the snippet candidate, the score of the snippet pattern applied to extraction of the snippet candidate, The weight for the snippet candidate may be calculated in consideration of one or more of the sentence lengths of the snippet candidates.

The snippet extracting step may be a semantic analysis of the snippet candidate, and may calculate the relevance of each element included in the snippet candidate to the query domain based on the relational dictionary.

Further, the search result providing step may include a template extracting step of extracting a template from the template DB based on the number of extracted snippets and the information type included in the snippet; And a search result generating step of matching the snippet with the template to generate a search result.

Further, the search result providing step may further include a snippet processing step of collecting and integrating a plurality of snippets.

Preferably, the search result providing step is a step of providing a search result by sequentially sorting a plurality of snippets, exposing a part of each snippet in correspondence with a predetermined size, The contents can be exposed.

According to another aspect of the present invention, there is provided a search service apparatus comprising: a query receiving unit receiving a query of a user; A snippet extractor for extracting a main content associated with the query from one or more documents retrieved based on the query, extracting a main content associated with the query, and extracting a snippet; And a search result providing unit for providing a search result for the query including the extracted snippet.

The snippet extracting unit preferably extracts a snippet pattern in which a content development structure corresponding to the query is defined and determines whether the snippet pattern is applied And extract the snippet from the document.

A natural language processor for extracting elements by morpheme analysis of documents or sentences; And a relational dictionary storing information on at least one of domain-specific words, dogs, abbreviations, synonyms, synonyms, or stereotypes.

Preferably, the query analyzing unit may include a query type analyzer for extracting an element from the query using the natural language processor and the related dictionary, and analyzing the query type based on the extracted element to determine a query domain .

The snippet extracting unit may collect the documents, analyze the elements and the structure in the document collected using the natural language processor and the related dictionary, determine the association between the elements, A snippet pattern generator for generating a snippet pattern by combining structures; A snippet candidate extractor for extracting a snippet candidate by analyzing a document retrieved on the basis of the elements included in the query by elements and structure using the natural language processor and the related dictionary, and applying a snippet pattern; And a snippet filter for extracting a snippet through morphological analysis and semantic analysis for the snippet candidate.

The snippet pattern generator further includes a snippet pattern DB for determining a snippet domain for the generated snippet pattern and storing the snippet pattern generated by the snippet pattern generator in association with the snippet domain can do.

Further, the query analyzing unit may receive the query of the user from the search service server, and the search result providing unit may provide the search result to the search service server.

Here, the snippet extracting unit may collect a document in a document DB storing a document in advance or on the Internet.

The search service apparatus according to the present invention may be a search service apparatus in which a computer program for performing each step of the search service providing method is recorded.

According to the present invention, a query and a document analysis of a user are performed through a natural language processing technique, and a pattern corresponding to a user's query is applied to a document to provide a result of summarizing contents related to a query in a document, The entire contents can be grasped.

In addition, by identifying domain specific content development patterns in advance and setting them as snippet patterns, and by applying the domain's snippet pattern corresponding to the user's query to the document, And can be extracted by a snippet.

In addition, according to the present invention, even if a document having a complicated structure is used, various types of snippet patterns can be applied to easily find corresponding contents in a document according to respective snippet patterns, and by integrating them, It is possible to extract a snippet that has been appropriately summarized.

Furthermore, when a search result screen is provided so that the user can grasp the search results including a large number of snippets, a plurality of snippets are associated with predetermined sizes to expose a part of each snippet, By exposing the entire contents of the selected snippet, the user can more easily grasp the search results containing multiple snippets at a glance.

1 shows an example of a search result through a search service according to the related art,
2 shows a schematic configuration diagram of a search service system for providing a query response type smart search service according to the present invention,
3 shows a configuration diagram of an embodiment of a search service apparatus according to the present invention,
4 shows a detailed configuration diagram of a main configuration of a search service apparatus according to the present invention,
Figure 5 shows a schematic flow diagram of an embodiment of a method of providing a query response type smart search service according to the present invention,
6 is a flowchart of an embodiment of a query analysis process in a query response type smart search service providing method according to the present invention,
FIG. 7 shows an embodiment of a query analysis process in a query response smart search service providing method according to the present invention,
8 shows a flowchart of an embodiment of a snippet pattern generation process in a method of providing a query response type smart search service according to the present invention,
9 shows an embodiment of a process of generating a snippet pattern in a method of providing a query response type smart search service according to the present invention,
10 is a flowchart illustrating an embodiment of a snippet candidate extraction process in a query response type smart search service providing method according to the present invention,
11 and 12 illustrate an embodiment of a snippet candidate extraction process in a query response type smart search service providing method according to the present invention,
13 shows a flowchart of another embodiment of a snippet candidate extraction process in a query response type smart search service providing method according to the present invention,
FIG. 14 shows another embodiment of a snippet candidate extraction process in a query response type smart search service providing method according to the present invention,
15 is a flowchart illustrating an embodiment of an integrated process of extracting a snippet pattern and extracting a snippet candidate in a query response type smart search service providing method according to the present invention,
16 is a flowchart of an embodiment of a snippet extraction process in a query response type smart search service providing method according to the present invention,
17 shows a flowchart of an embodiment of a search result generation process in a query response type smart search service providing method according to the present invention,
18 shows an embodiment of a search result providing process in a query response type smart search service providing method according to the present invention,
FIG. 19 shows another embodiment of a search result providing process in a query response type smart search service providing method according to the present invention,
20 shows another embodiment of the search result providing process in the query response type smart search service providing method according to the present invention,
FIG. 21 shows a comparative example of a conventional search service and a search service according to the present invention.

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 of the present invention when taken in conjunction with the accompanying drawings.

First, the terminology used in the present application is used only to describe a specific embodiment, and is not intended to limit the present invention, and the singular expressions may include plural expressions unless the context clearly indicates otherwise. Also, in this application, the terms "comprise", "having", and the like are intended to specify that there are stated features, integers, steps, operations, elements, parts or combinations thereof, But do not preclude the presence or addition of features, numbers, steps, operations, components, parts, or combinations thereof.

In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

The present invention recognizes a query through natural language processing on a query input by a user, exposes snippet information including the overall contents of the query by organizing the contents of a document related to the query into a corresponding pattern, This paper proposes a smart search service that can easily grasp the specific contents required by users on the document.

FIG. 2 shows a schematic diagram of a search service system for implementing a query response type smart search service according to the present invention.

The search service system may include a user terminal 10, a search service apparatus 100, and a document 50, and may further include a separate search service server 30 depending on the situation.

The user terminal 10 is a communication terminal for inputting and receiving a query from a user and for receiving a search result of a query and providing the search result to a user, and includes a wired / wireless communication terminal commonly used such as a PC, a notebook computer, a PDA, And may be a specialized communication terminal that is individually provided to provide a search service according to the present invention.

The search service apparatus 100 is a device for providing a search service by generating a query response type smart search result with respect to a query of a user, and may be configured as a single server or may be a combination of a plurality of devices or modules, Group. Here, the query-response type smart search result is different from the existing search result that only provides a certain region in which a word used in a query exists in the document as a search result, and the user's query and contents of the document are analyzed based on a natural language processing technique And summarizes the contents related to the query in the document, and includes the snippet containing the overall contents of the query.

Also, the user terminal 10 can access the search service apparatus 100 through the network and receive the search service. The network can include a wired network and a wireless network as general Internet lines. Further, A person may include a virtual private network or an intranet accessible under certain conditions.

The document 50 includes not only documents such as information posted on blogs and cafes that can be easily collected by anyone via the Internet 50a but also papers or business documents provided through a specialized database providing services to a limited member or a predetermined area In the present invention, a document is used to mean not only plain text but also all data including information such as image, sound, image and the like.

Furthermore, a document DB 50b, which collects and stores various documents in advance through a web bot or the like, may be constructed, and the search service apparatus 100 may provide a search service based on a document held on the document DB 50b .

In addition, the search service apparatus 100 according to the present invention may provide a query response type smart search service according to the present invention in cooperation with a search service server 30 that provides a general search service. For example, When the user accesses the search service server 30 and inputs a query to the search service server 30, the search service server 30 requests the search service apparatus 100 according to the present invention to send a query response type smart search result for the user query The search service apparatus 100 may generate a query response type smart search result for the user's query, and the result may be provided to the user through the search service server 30. [

A search service apparatus for providing a query response type smart search service according to the present invention will be described with reference to an embodiment shown in FIG.

The search service apparatus 100 according to the present invention may include a query analyzing unit 110, a snippet extracting unit 130 and a search result providing unit 150 in addition to the related language dictionary 210, A nipple pattern DB 250, a template DB 270, and the like. 3, a query receiving unit for receiving a query from the user terminal 10 may be included in the search service apparatus 100. [ In addition to the basic configuration, additional configurations may be included in the search service apparatus 100, but they may be configured as separate apparatuses depending on the circumstances, and may be interworked with the respective configurations of the search service apparatus 100.

The query analyzing unit 110 analyzes the query of the user to determine the query type and determines the query domain according to the query type. For example, the query analyzer 110 extracts the elements included in the query using the related word dictionary 210, And the query domain is determined through the analysis of the relationship between the elements and the elements. The query analyzing unit 110 may directly provide a search service page to the user terminal 10 to receive a query from the user terminal 10 or may interwork with a separate search service server 30 or through a separate query receiving unit The user's query may be sent.

The snippet extractor 130 extracts a snippet pattern corresponding to the query and extracts the snippet on the document 50 related to the query by applying the snippet pattern.

A snippet pattern is a structure in which contents are developed in a document. In general, a document has a certain structure and contents are presented according to the structure. In particular, the structure of documents may be different depending on a specific domain have. Accordingly, if the structural form of the document is identified, the main contents desired in the document can be easily extracted and arranged. In the present invention, the structural form in which the contents are developed in the document is defined and set as a snippet pattern.

The snippet pattern database 250 stores the snippet patterns set by the administrator in each domain in advance or the snippet extractor 130 collects the various documents 50 and collects the contents on the document based on the natural language processing technique And the structure thereof is recognized, and the structure type in which the main contents are presented on the document is recognized based on the structure, and it can be generated as a snippet pattern and stored in the snippet pattern DB 250.

The snippet extracting unit 130 retrieves and extracts a snippet pattern corresponding to the query domain from the snippet pattern DB 250 and applies the extracted snippet pattern to the document 50 related to the query to extract the structure of the snippet pattern Depending on the form, you can extract the snippet by extracting the main contents from the document.

In the present invention, a snippet does not simply mean only one minute containing a specific word but refers to information summarizing the overall contents of the document related to the query by extracting contents related to the query on the document and processing the collected information.

The snippet pattern and the snippet presented in the present invention will be described in more detail in the following embodiments.

The search result providing unit 150 provides a search result of the user's query based on the snippet extracted by the snippet extracting unit 130. The search result providing unit 150 may provide a search result in which the extracted snippet is directly arranged, The information structure included in the snippet may be identified, the corresponding template may be extracted from the template DB 270, and the retrieved result may be generated by matching the snippet with the extracted template.

Further, the search result providing unit 150 exposes only a part of the plurality of snippets according to the screen size when providing the search result screen including a plurality of snippets, and exposes the entire contents of the selected snippet according to the user's selection A search result screen may be provided.

4, the search service apparatus 100 may have a natural language processor 120 for more effectively performing query analysis and document analysis, The query analyzing unit 110 and the snippet extracting unit 130 extract the elements by morphological analysis of the query or document through the natural language processor 120 and use the related language dictionary 210 to perform semantic analysis Analyze the association.

Here, the relative word dictionary 210 may be a simple term dictionary, but it is preferable to selectively retain information such as domain-specific words, words, idioms, synonyms, synonyms or stereotype according to need, This can be a tagged, comprehensive dictionary of meaning.

The query analyzer 110 includes a query type analyzer 115 and the query type analyzer 115 extracts elements from the user's query using the natural language processor 120 and the dictionary dictionary 210, And determines the query domain by analyzing the query type.

The snippet extracting unit 130 may include a snippet pattern generator 131, a snippet candidate extractor 133, a snippet filter 135, and the like.

The snippet pattern generator 131 analyzes elements and structures in a document by using the natural language processor 120 and the dictionary dictionary 210, determines an association between the elements, The structure is combined to create a snippet pattern. Although the snippet pattern extractor 133 may directly use the snippet pattern generated by the snippet pattern generator 131, the generated snippet pattern is preferably stored in the snippet pattern DB 250, The pattern generator 131 determines the snippet domain for the snippet pattern and stores the generated snippet pattern in the snippet pattern DB 250 in correspondence with the corresponding snippet domain.

The snippet candidate extractor 133 analyzes the document retrieved based on the elements included in the query by using the natural language processor 120 and the dictionary 210 in the element and structure, Extract the snippet candidate. If the snippet pattern DB 250 is constructed in advance, the snippet candidate extractor 133 extracts a snippet pattern corresponding to the query domain analyzed by the query analyzer 110 from the snippet pattern DB 250 You can extract the snippet candidates by extracting them and applying them to the document.

The snippet filter 135 filters the snippet candidates extracted by the snippet candidate extractor 133 to extract snippets suitable for the query. The snippet candidates are filtered through shape analysis and semantic analysis of the snippet candidates You can extract the snippet.

The search service apparatus according to the present invention can implement the query response type smart search service by having such configurations, and each of the configurations may be configured as an individual device and a module, or may be included in one device. Furthermore, the search service apparatus according to the present invention may be replaced with a server of a kind. For example, a computer program implementing a query response type smart search service providing method according to the present invention to be described later is installed in a server, .

The present invention proposes a method of providing a query response type smart search service using the search service apparatus described above. Hereinafter, a method of providing a query response type smart search service according to the present invention will be described with reference to the embodiments. do. The query response type smart search service providing method according to the present invention is implemented through a search service system including the search service apparatus according to the present invention described above, and will be described with reference to the same.

A method of providing a query response type smart search service according to the present invention includes: a query receiving step of receiving a query of a user; A document retrieval step of retrieving a document based on the query; A snippet extracting step of extracting a main content associated with the query from at least one of the retrieved documents, extracting a main content associated with the query, and extracting a snippet; And a search result providing step of providing a search result for the query including the extracted snippet.

FIG. 5 shows a schematic flow diagram of an embodiment of a method of providing a query response type smart search service according to the present invention.

5 may include a query analysis process S100, a snippet extraction process S300, and a search result providing process S400. Each process may be performed in a manner that does not deviate from the main technical features of the present invention Can be selectively changed as needed. FIG. 5 briefly explains the concept of each process, and details of each process will be described in detail with reference to the following embodiments and implementation examples.

In the query analysis process S100, when a query from the user is input (S110), the query type for the query is analyzed based on the natural language processing technique (S150) to determine the query domain. The query domain according to the query type is classified in advance, and in the query analysis process (S100), it is determined which query domain corresponds to the user query.

In the snippet extraction process (S300), a snippet pattern corresponding to the domain determined through the query analysis process (S100) is extracted (S310). At this time, if the snippet pattern DB , The snippet pattern corresponding to the query domain can be extracted from the snippet pattern DB 250. [ If the snippet pattern DB 250 is not applied, the snippet pattern can be extracted by analyzing the document related to the query based on the natural language processing technique. Then, the extracted snippet pattern is applied to the document related to the query to extract the snippet candidate (S330), and the extracted snippet candidate is filtered to extract an appropriate snippet in response to the query (S350).

As described above, the snippet pattern is a structure in which the main content is presented on the document. In the present invention, the snippet pattern is extracted and the snippet pattern is applied to the document related to the query, And extracts the snippet candidates. Then, the extracted snippet candidate is selected by filtering whether the search result is appropriate for the query, and the selected snippet candidate is extracted as a snippet for the query.

Next, the search result providing step (S400) generates a search result for the query based on the extracted snippet (S410) and provides it as a response to the user's query. If necessary, the user can easily search results You can integrate or manipulate snippets to provide insight into your search results.

The method of providing a query response type smart search service according to the present invention performs user query and document analysis through a natural language processing technique and applies a pattern corresponding to a user query to a document, So that the entire contents can be grasped only by the search result.

The details of each process of the query response smart search service providing method according to the present invention will be described in more detail.

First, with reference to a query analysis process S100, FIG. 6 shows a flowchart of an embodiment of a query analysis process in a query response type smart search service providing method according to the present invention.

The query analyzing unit 110 analyzes the query of the user through the natural language processor 120 and the related dictionary 210 to extract the element from the query S151 and determines the clue element and the keyword element from the extracted element S153). Then, the query type according to the combination of the clue element and the keyword element is analyzed (S155), and the query domain corresponding to the query type is determined (S157).

For example, referring to the embodiment of the query analysis process in the query response type smart search service providing method according to the present invention shown in FIG. 7, When the 'kimchi stew recipe' is inputted as the query of the 'kimchi stew recipe', it is analyzed through the natural language processor 120 and the dictionary dictionary 210, and the 'recipe' The 'Kimchi stew' may be extracted as the keyword element 313 corresponding to the clue element.

This is because the domain in which the clue element 311 and the keyword element 313 correspond can be determined on the basis of the domain-specific word held in the related word dictionary 210. At this time, It is possible to determine whether the keyword element 313 is a domain word corresponding to the domain of the clue element 311 after the domain of the clue element 311 is first recognized.

If the meaning and the tag for the domain element 311 and the keyword element 313 for the clue element 311 are grasped through the relation dictionary 210 as shown in FIG. 7C, The combination of the clue element 311 and the keyword element 313 is classified into a food name and a food item by the combination of the clue element 311 and the keyword element 313, The query type for the recipe is grasped, and the domain for the query can be judged as a 'recipe' as shown in FIG. 7 (d).

As described above, in the query analysis process (S100), the domain of the query is determined by analyzing the query type through the natural language processing technique and the related dictionary having the domain-specific words defined in advance.

Next, with respect to the process of extracting the snippet pattern (S300), a process of constructing the snippet pattern DB will be described with reference to the flow chart of the embodiment of the snippet pattern generation process in the query response type smart search service providing method according to the present invention shown in FIG. As shown in FIG.

First, when a web bot or the like searches for a document (S210) and collects the collected document, the snippet extractor 130 analyzes the collected document as a respective element through the natural language processor 120 (S220). Here, the analysis of the elements of the document can be performed through morphological analysis through the natural language processor 120, sentence separation, and the like.

The sniff domain is determined by determining the clue element and the corresponding keyword element in the analyzed element by using the related word dictionary 210 and determining the domain according to the combination of the clue element and the keyword element. And the process of determining the snippet domain is omitted in the flowchart of FIG. 8 because the order of the other processes may be changed depending on the situation.

When the element analysis of the document is completed, the structure of the region including the element is analyzed (S230), the relationship between the elements is determined (S240), and the analyzed structure is combined And creates a nipple pattern (S250).

At this time, the snippet pattern can be generated by distinguishing the domain-specific specialized snippet pattern corresponding to the specific domain and the common snippet pattern corresponding to all the domains. For example, a simple sequential list, a table list, List can be divided into common snippet pattern, and the combination of structure such as material list and cooking order list can be applied to the recipe domain in a limited manner, so that the specialization snippet of the recipe domain Patterns. The distinction between the common snippet pattern and the specialized snippet pattern can be changed according to the condition set by the administrator, and the administrator can arbitrarily classify the extracted snippet pattern into a common snippet pattern or a specialized snippet pattern.

The generated snippet pattern is stored in the snippet pattern DB 250, and the generated snippet pattern is stored in association with the snippet domain determined in advance.

Referring to FIG. 9, which is an embodiment of generating a snippet pattern, an element is analyzed based on a natural language processing technique for the document 410 shown in FIG. 9A, and a clue element and a keyword element There is a 'recipe' 413 as a clue element and 'oil pasta' 412 exists as a keyword element corresponding to the 'recipe' 413. Therefore, in the case of the document shown in FIG. 9A, the domain for the document 410 can be classified as 'recipe' according to the combination of the clue element and the keyword element. This is similar to the process of determining a query domain through the above-described FIG. 7, and thus a detailed description thereof will be omitted.

When the structure of the document 410 is analyzed based on the analyzed elements, there are a recipe 413, a material 413, a cooking method 415 and the like as main elements, and a recipe 413 , The material words associated with the material 413 are listed 414 in the stop area and the cooking order list 416 associated with the cooking method 415 is listed in the bottom area Respectively. The material words 413 are related to the area where the 'material' 413 corresponds to the 'oil pasta' and the 'recipe' included in the title 411, The cooking order list 416 may be analyzed as being associated with the area where the 'cooking method' 415 is located corresponding to the 'oil pasta', 'recipe' and 'material' 413 included in the title 411 have. A snippet pattern is generated by combining structures based on this association analysis.

For example, when the analysis result of the document 410 shown in FIG. 9A is integrated, as shown in FIG. 9B, the snippet domain is a 'recipe' and the 'order list' , And a specialized snippet pattern 430 composed of a title, a material, and a cooking method corresponding to the recipe domain can be generated.

In addition, when the analysis results of the document 410 shown in FIG. 9A are individually separated, two types of snippet patterns can be generated as shown in FIG. 9C, The specialized snippet pattern 450 composed of the title and the material corresponding to the recipe domain can be generated as the 'recipe' and the 'list entity name' pattern 440 for the material, Recipe 'and the' order list 'pattern 460 for the cooking order, a special snippet pattern 470 corresponding to the recipe domain and composed of title and cooking order may be generated.

Through this process, various snippet patterns can be extracted and stored in the snippet pattern DB 250.

FIG. 10 is a flowchart illustrating a method for extracting a snippet candidate for a query by applying a snippet pattern in a snippet extracting process (S300). FIG. 10 is a flowchart illustrating a process of extracting a snippet candidate from a query response- 1 shows a flow chart of an embodiment of the present invention.

The snippet extracting unit 130 searches the snippet pattern DB 250 for a snippet pattern corresponding to the query domain for the query of the user determined through the embodiment of FIG. 6 or the like (S311) , Searches the snippet domain corresponding to the query domain in the snippet pattern DB 250, and extracts a snippet pattern stored in a matching manner with the retrieved snippet domain.

In addition, the snippet extracting unit 130 searches for a document based on the elements included in the query of the user (S331). At this time, the document may be searched through the Internet or the like, if necessary, You may.

The snippet extracting unit 130 analyzes the searched document as an element by using the natural language processor 120 and the dictionary 250 in operation S332 and applies the extracted snippet pattern corresponding to the query domain to the searched document, The snippet candidate is extracted on the document (S333).

As described in the embodiment of the above 9, the snippet pattern is a form having contents on the document. The snippet pattern is applied to extract the contents from the document according to the snippet pattern, The snippet candidate can be extracted.

Referring to FIG. 11, FIG. 11 (a) illustrates a query process 510 of a user, as a result of a query analysis process, It is an example of judging the element and the query domain 515 'fitness' by analyzing based on natural language processing technology.

11 (b) is searched based on the elements included in the query, and a process of extracting a snippet candidate from the searched document 530 will be described. First, in the retrieved document 530, And performs a natural language process on the document 530 to analyze each element on the document 530.

Then, the snippet pattern 530 extracted from the snippet pattern DB 250 is applied to the document 530 in correspondence with the query domain 515 'fitness', where the snippet domain corresponding to the query domain 'fitness' Fitness ", and a snippet pattern 530 composed of a title and a movement order list is extracted and applied.

Since the snippet pattern 530 is composed of a title and a movement order list and is applied to the document 530, the element 531 indicating the title and the elements 533 and 534 indicating the order on the document are included in the snippet pattern 530 The contents 532 and 535 including the contents 533 and 534 indicating the contents and the contents including the element 531 indicating the title can be found out.

The snippet candidate 540, which summarizes the content of the title and the order of the exercise, is extracted on the document 530 by collecting and processing the excavated contents according to the snippet pattern 530. [

Referring to FIG. 12, (a) of FIG. 12 shows a result of a query analysis process, a query 550 of the user ' And analyzed based on a natural language processing technique to determine an element and a query domain 555 as a 'moving picture list'. A process of extracting a snippet candidate from the retrieved document 570 will be described with reference to the document 570 shown in FIG. 12B based on the elements included in the query.

A natural language process is performed on the searched document 570 to analyze each element on the document 570 and the snippets extracted from the snippet pattern DB 250 corresponding to the query domain 555 ' The pattern 560 is applied to the document 570 where a snippet pattern 560 consisting of an ordered list of titles, video names, and URLs corresponding to the query domain 'video list' And applies it to the document 570.

On the document 570, there are the elements 571 and 573 indicating the moving picture name and the moving picture element 572, and the moving picture element 572 exists after the element 571 indicating the moving picture name sequentially. Therefore, when the form of the snippet pattern 560 is applied, the element 571 representing the moving picture name and the link URL of the moving picture element 572 can be found in association with each other on the document 570.

The snippet candidate 580 in which the contents combined with the moving picture name and the moving picture URL are summarized on the document 570 is extracted by collecting and processing the excavated contents according to the snippet pattern 560 and processing.

Further, in the process of extracting a snippet candidate for a query by applying a snippet pattern in the snippet extraction process (S300), a plurality of snippet candidates are extracted by applying a plurality of snippet patterns to one document, 13 is a flow chart of another embodiment of the snippet candidate extracting process in the query response type smart search service providing method according to the present invention. do.

The embodiment of FIG. 13 shows a process after extracting a plurality of snippet patterns (S312) in the embodiment of FIG. 10 and analyzing the retrieved document based on the natural language processing technique (S332).

The snippet extracting unit 130 extracts the snippet candidates by applying each of the plurality of snippet patterns to the analyzed document S335. In FIG. 13, the common snippet pattern is applied to the analyzed document, (Step S336), and extracting a snippet candidate by applying a specialized snippet pattern (step S337). However, according to the extracted snippet pattern, a plurality of specialized snippet patterns or a plurality of common snippet patterns May be applied.

Then, a plurality of snippet candidates extracted by applying each of the plurality of snippet patterns are collected and integrated (S338), whereby a plurality of snippet patterns are applied to one document to extract the snippet candidates (S339).

14, another embodiment of the snippet candidate extraction process in the method of providing a query response type smart search service according to the present invention is shown.

In the description of the embodiment of FIG. 14, a process of extracting elements by analyzing a document and analyzing the entire structure through area analysis has been described in detail with reference to FIGS. 11 and 12, The concept of applying a pattern is briefly described.

The embodiment of FIG. 14 is a case of extracting a snippet candidate for a query of the 'pasta cooking method' of a user, wherein a snippet domain is a 'recipe' and a specialized snippet pattern for a 'recipe' 620) and a common snippet pattern 640 composed of an order list applicable to all domains are extracted and applied to the document 610.

Applying a specialized snippet pattern 620 for a 'recipe' consisting of a title and a material to a document 610 causes a snippet candidate 630 for the content of the material list according to the specialized snippet pattern 620 on the document 610 Can be extracted. In addition, when a common snippet pattern 640 composed of an order list is applied to the document 610, the snippet candidates 650 for the contents listed in a sequential order on the document 610 can be extracted. In this case, The cooking sequence can be extracted corresponding to the snippet pattern as the listed contents.

If the snippet candidate 630 extracted by applying the specialized snippet pattern 620 and the snippet candidate 650 extracted by applying the common snippet pattern 640 are integrated and processed, A snippet candidate 660 is extracted which summarizes the contents of the title, the material, and the cooking order.

As described above, according to the present invention, by applying various types of snippet patterns to one document, it is possible to easily find each content according to each snippet pattern even in a document having a complicated structure. And can summarize appropriately.

Further, the snippet extraction process (S300) can extract a snippet candidate by extracting a snippet pattern on a document retrieved in association with a query without directly constructing a separate snippet pattern DB, FIG. 15 is a flowchart illustrating an embodiment of a process of integrating a snippet pattern extraction and a snippet candidate extraction in a query response type smart search service providing method according to the present invention.

FIG. 15 is a process after the analysis of the query type of the user is completed through the query analysis process (S100). The document is searched (S315) based on the elements included in the query of the user (S315) The retrieved document is analyzed (S316). Then, the analyzed structure of the element area is analyzed (S317), the overall shape of the document is identified, and the relation between the elements is determined (S318).

Through the analysis process of the document, a snippet pattern corresponding to the query type is extracted (S319) based on the relation between the overall structure of the document and the element. 15, the process of analyzing a document and extracting a snippet pattern from a document is similar to the embodiment of FIG. 8 described above. In FIG. 15, a point of extracting a snippet pattern from a retrieved document 8 is different from the above.

After extracting the snippet pattern from the retrieved document in association with the query of the user, the snippet pattern is applied to the document immediately and the snippet candidate is extracted on the document (S335). The extraction of the snippet candidates has been described in detail with reference to the embodiments of FIGS. 11, 12 and 14, and thus a detailed description thereof will be omitted.

Next, in the process of extracting the snippet from the snippet candidates in the snippet extracting process (S300), only one or more snippet candidates are extracted, and only appropriate snippet candidates are selected through filtering to extract them as snippets. 16 is a flowchart illustrating an embodiment of a snippet extraction process in a query response type smart search service providing method according to the present invention.

In the state (S351) in which the snippet candidates related to the user's query are extracted and prepared through the various embodiments and the implementation examples described above, the snippet extractor 130 extracts the snippet candidate from the form analysis (S352) (S354), an appropriate snippet is extracted in response to the query.

A form analysis (S352) for a snippet candidate includes: an application count for a snippet pattern applied to extraction of the snippet candidate; an evaluation score for a snippet pattern applied to the extraction of the snippet candidate; The weight of the snippet candidate is calculated in consideration of various external factors of the snippet candidate such as the length (S353).

For example, an evaluation of a snippet pattern applied to extract a snippet candidate with respect to a user query can determine whether the extracted snippet candidate is an appropriate response to the query. A plurality of snippet candidates The more the one specific snippet pattern is applied, the higher the relevance of the query and the specific snippet pattern is. Therefore, the snippet candidate extracted by applying the specific snippet pattern may be expected to be suitable for the query have. In addition, when a snippet candidate extracted with a specific snippet pattern is extracted as a snippet and provided as a search result, the user's response to the snippet is determined, or the reliability of the specific snippet pattern is set in advance by the manager The evaluation score for the nipple pattern may be calculated and the evaluation score calculated in the snippet pattern may be taken into consideration to determine whether the snippet candidate applied to the query is suitable for the query.

Furthermore, if the sentence length of the snippet candidate is too short or too long, it is highly probable that the content of the snippet candidate deviates from the query. Therefore, the sentence length information of the snippet candidate may be considered.

In consideration of various external factors of the snippet candidate, the weight for the snippet candidate may be calculated (S353) through the type analysis of the snippet candidate.

As a semantic analysis (S342) for a snippet candidate, the relevance degree of each element included in the snippet candidate to the query domain is calculated on the basis of a related word dictionary.

For example, the more snippets related to the query domain are included in the content of the snippet candidate, the more likely it is that the snippet candidate is an appropriate response to the query. Therefore, the snippet candidate based on the domain- And the relevance to the snippet candidate is calculated (S355).

In step S356, the snippet candidates are filtered based on the weights and related degrees calculated for the snippet candidates. For example, a function for assigning the weights and the degrees of association to the respective variables is set, and the weights and the degrees of association And if the degree of the result is within the set range, the snippet candidate may be selected as a snippet (S257).

When the snippet is extracted through the above process, the extracted snippet is collected to generate a search result for the user's query. In this case, the search result obtained by simply listing the extracted snippet may be provided, The extracted snippet can be processed in a form that can be easily recognized by the user, thereby providing search results. As an example, the content contained in a snippet can be easily recognized, structured, or tabulated by a user based on a related dictionary, and furthermore, a plurality of snippets can be integrated into a single snippet have.

FIG. 17 shows a flowchart of an embodiment of a search result generation process in a query response type smart search service providing method according to the present invention, as an example of a search result providing process (S400).

The search result generation unit 150 comprehensively determines the number of the snippets to be provided as a search result for the query and the information types included in the snippet in step S411, The template corresponding to the information type of the snippet is extracted (S412), the snippet is matched to the extracted template type (S413), and the search result is generated (S414).

18 shows an embodiment of a search result providing process in a query response type smart search service providing method according to the present invention.

18A, a simple text format template 815 corresponding to the information format included in the snippet 810 is extracted from the template DB 270 in the case of the simple text type snippet 810 And the search result is generated by matching the snippet 810 with the extracted template 815. And provides a screen 820 containing the generated search results to the user.

In the case of the snippet 820 in which the text and the image are mixed as shown in FIG. 18 (b), the template 825 having the text and image arrangement positions set therein is extracted from the template DB 270, The snippet 820 is matched to generate a search result. And provides a screen 830 containing the generated search results to the user.

In the embodiment of FIG. 18, since one snippet itself is directly matched to a template to generate a search result, no special processing is required for the snippet. However, when a plurality of snippets are collected, In this regard, FIG. 19 shows another embodiment of the search result generation process in the query response smart search service providing method according to the present invention.

The content 851 and 853 included in one snippet 850 and the content 861 included in the other snippet 860 in the aggregation of the two snippets 850 and 860 , 863), it is possible to grasp the duplicated contents 853, 861 by comparing the contents included in each of the snippets 850, 860 with each other, collect the duplicated contents, The kneaders 850 and 860 are processed. Then, a snippet 840 is created from the merged contents 870 by combining the contents of the two snippets 850 and 860 through the processing.

When the processing for the snippet is completed, the template 875 corresponding to the processed snippet 840 is extracted from the template DB 270, and the snippet 840 is matched with the extracted template to generate a search result.

Furthermore, a search result 880 including a plurality of snippets 881, 883, and 885 for one query may be generated and provided as a search result screen.

As shown in FIG. 19, when providing search results 880 including a plurality of snippets 881, 883, and 885 for one query, search results including a plurality of snippets 881, 883, and 885 It is possible to expose only a part of the contents of each of the snippets 881, 883 and 885 to the search result screen so that the user can recognize the search result 830 at a time. In this regard, FIG. 2 shows an embodiment of a search result providing method in a search method.

20A, a plurality of snippets 911 and 915 are sequentially arranged and provided in the search result screen 910, and only a part of the contents of each of the snippets 911 and 915 is displayed on the screen And places a button 912 on the screen that allows selection of the entire contents of the snippet 911. [

When the user clicks a button 912 for selecting the entire contents appearance of the snippet 911, a search result screen (not shown) is displayed so that the entire contents of the selected snippet 921 are exposed as shown in FIG. 920) are reconstructed and provided. A button 922 for selecting a portion of the snippet 921 in which the entire content is exposed is displayed again. When the snippet 921 is selected, the search result screen 910 shown in FIG. 20A is reconstructed / RTI >

As described above, in the present invention, a snippet summarizing the main contents of a document is provided as a search result, so that one snippet occupies a large area on the screen and can be exposed. Accordingly, in order to allow users to grasp search results including a large number of snippets, a plurality of snippets are corresponded to predetermined sizes to expose a part of each snippet, and the entire contents of the snippet By exposing, the user can more easily grasp search results containing a large number of snippets at a glance.

As described above, the query response type smart search service according to the present invention provides a result of summarizing the main contents of a document related to a query, so that the entire contents can be grasped only by a search result. Service and a search service according to the present invention.

As shown in FIG. 21, the search result screens 930 and 940 according to the conventional search service merely show only a part of the keyword as a search result centering on the keywords included in the user's query. Therefore, If you do not know the answer and click on the contents included in the search result to access the document, you will have to understand the whole contents. Especially, when you access the document with only contents included in the search result, The case is common.

However, since the search results screen 940 and 950 according to the query response type smart search service according to the present invention are summarized as the search result as a summary of the main contents included in the document related to the query, the user can search for the corresponding document You will be able to see the overall contents of the document, and the search results will give you an overall understanding of the document so that you can find the document you want more easily and quickly.

Further, the query response type smart search service according to the present invention may include a search result screen 940, which includes a search result 941 including a snippet according to a query response type search service and a search result 942 including a snippet according to a conventional search service Only the uppermost snippet having high accuracy when providing the search result screen 940 is extracted and provided by the search service providing method according to the present invention, and the remaining remaining snippets are provided to the existing search service providing method It is possible to provide more search results to the user while providing accurate search results.

Further, the search result screen 950 may be configured to include only the snippets 951 and 955 according to the query response type search service, and a snippet according to the query response type search service according to the present invention and an existing search service You can also configure a search results screen with a mix of different snippets. Such a search result screen can be changed according to a user's selection or an administrator's setting.

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.

10: user terminal,
30: Search service server,
50: Document
100: Search service device,
110: query analysis unit, 115: query type classifier,
120: natural language processor,
130: a snippet extracting unit, 131: a snippet pattern generator,
133: snippet candidate extractor, 135: snippet filter,
210: Relational dictionary,
250: Snippet pattern DB,
270: Template DB.

Claims

Receiving a query of a user;
A query type analysis step of analyzing a query type of the query to determine a query domain;
Extracting a snippet pattern defining a content expansion structure corresponding to the query domain from at least one of documents retrieved based on the query, applying the snippet pattern to the document, A snippet extracting step of extracting a main content and extracting a snippet; And
And a search result providing step of providing a search result of the query including the extracted snippet.

delete

The method according to claim 1,
In the snippet extracting step,
A snippet candidate extracting step of extracting a snippet candidate by extracting main contents from the document according to the snippet pattern by applying the snippet pattern to the document, and extracting a snippet candidate; And
And a snippet selecting step of selecting a snippet from the snippet candidate through a type analysis and a semantic analysis of the snippet candidate.

The method according to claim 1,
Wherein the query type analysis step comprises:
A query determination step of extracting an element from the query and determining a clue element and a keyword element corresponding to the clue element in the element through a relation dictionary;
A query type determination step of determining a query type according to the combination of the clue element and the keyword element; And
And a domain determination step of determining a query domain corresponding to the query type.

The method of claim 3,
And a snippet pattern database building step of building a snippet pattern DB storing a domain specific snippet pattern,
The snippet candidate extracting step includes:
And extracting a snippet pattern corresponding to the query domain from the snippet pattern DB.

6. The method of claim 5,
The step of constructing the snippet pattern DB comprises:
A document collecting step of collecting a document, performing a natural language process on the collected document, and analyzing the document into respective elements;
Determining a clue element and a keyword element corresponding to the clue element in the element through a previously held relation dictionary, and determining a snippet domain according to the combination of the clue element and the keyword element;
A document analysis step of analyzing a structure of an area including the element;
A snippet pattern generation step of determining an association between the elements and generating a snippet pattern by combining the analyzed structures based on the association; And
And storing the snippet pattern in the snippet pattern DB in association with the snippet domain.

The method of claim 3,
The snippet candidate extracting step includes:
Searching for a document based on the elements included in the query, performing a natural language process on the retrieved document, and analyzing the natural language process with each element; And
And extracting a snippet candidate by extracting contents corresponding to the query from the document according to the snippet pattern and collecting the extracted snippet pattern in accordance with the snippet pattern. Delivery method.

The method of claim 3,
The snippet candidate extracting step includes:
Extracting a plurality of snippet patterns corresponding to the query domain, extracting a plurality of snippet candidates by applying each snippet pattern to one document,
Further comprising a snippet candidate processing step of collecting a plurality of extracted snippet candidates and consolidating them into one snippet candidate.

The method according to claim 6,
The snippet pattern generation step may include:
And generating a common snippet pattern corresponding to each domain by distinguishing a special snippet pattern for each domain corresponding to a specific domain and a common snippet pattern corresponding to all the domains.

The method of claim 3,
The snippet candidate extracting step includes:
Collecting a document associated with the query, and performing a natural language process on the collected document to analyze the document into respective elements;
Determining a clue element corresponding to the clue element and a keyword element corresponding to the clue element in the element through a previously held relation dictionary, determining a domain according to the combination of the clue element and the keyword element, and comparing the domain with the query domain;
A document analysis step of analyzing a structure of an area including the element;
A snippet pattern generation step of determining an association between the elements and generating a snippet pattern by combining the analyzed structures based on the association; And
And extracting a snippet candidate by extracting contents corresponding to the query from the document according to the snippet pattern and collecting the extracted snippet pattern in accordance with the snippet pattern. Delivery method.

The method of claim 3,
In the snippet extracting step,
A weight is calculated by a morphological analysis of the snippet candidate, a relevance degree is calculated by semantic analysis of the snippet candidate, and a snippet is extracted by filtering the snippet candidate based on the weight value and the degree of association The search service providing method comprising:

12. The method of claim 11,
In the snippet extracting step,
The method according to claim 1, wherein, as the type analysis of the snippet candidate, the number of times of application to the snippet pattern applied to the extraction of the snippet candidate, the evaluation score of the snippet pattern applied to the extraction of the snippet candidate, And a weight for the snippet candidate is calculated in consideration of one or more of the snippet candidates.

12. The method of claim 11,
In the snippet extracting step,
Wherein the degree of relevance of each element included in the snippet candidate to the query domain is calculated based on the relation dictionary as a semantic analysis of the snippet candidate.

The method according to claim 1,
The search result providing step may include:
Extracting a template from the template DB based on the number of extracted snippets and the information type included in the snippet;
And a search result generation step of matching the snippet with the template to generate a search result.

15. The method of claim 14,
The search result providing step may include:
And a snippet processing step of collecting and integrating a plurality of snippets.

The method according to claim 1,
The search result providing step may include:
Wherein a search result is provided by sequentially arranging a plurality of snippets, exposing a part of each snippet in correspondence with a predetermined size, and exposing the entire contents of a snippet selected at the time of a user's selection Delivery method.

A query receiving unit for receiving a user query;
A query analyzer for analyzing the query to determine a query type;
Extracting a snippet pattern in which a content development structure corresponding to the query is defined, extracting a main content associated with the query from one or more documents retrieved based on the query by applying the snippet pattern, A snippet extracting unit for extracting a snippet; And
And a search result providing unit for providing a search result for the query including the extracted snippet.

delete

18. The method of claim 17,
A natural language processor for extracting elements by morpheme analysis of documents or sentences; And
And a relation dictionary storing information on at least one of a domain-specific word, an entity word, an abbreviation, a synonym, a thesaurus, or a variant.

20. The method of claim 19,
The query analyzing unit,
And a query type analyzer for extracting an element from the query using the natural language processor and the related dictionary, and analyzing a query type based on the extracted element to determine a query domain.

20. The method of claim 19,
The snippet extracting unit extracts,
Analyzing elements and structures in a document collected by using the natural language processor and the related dictionary, judging an association between the elements, and combining the analyzed structures based on the association, A snippet pattern generator for generating a snippet pattern;
A snippet candidate extractor for extracting a snippet candidate by analyzing a document retrieved on the basis of the elements included in the query by elements and structures using the natural language processor and the related dictionary, and applying a snippet pattern; And
And a snippet filter for extracting a snippet through morphological analysis and semantic analysis of the snippet candidate.

22. The method of claim 21,
The snippet pattern generator comprising:
Determine the snippet domain for the snippet pattern you created,
And a snippet pattern DB for storing the snippet pattern generated by the snippet pattern generator in association with the snippet domain.

18. The method of claim 17,
The query analyzing unit,
Receives a query of the user from the search service server,
The search result providing unit,
And provides the search result to the search service server.

18. The method of claim 17,
The snippet extracting unit extracts,
And collects the document on the document DB or the Internet storing the document in advance.

A computer program stored in a computer-readable medium for causing a computer to execute the steps of the method of providing a search service according to any one of claims 1 to 17.