JP2010205265A - Content retrieval system and method using ontology - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To disclose a content retrieval method and a system which uses ontology. <P>SOLUTION: The content retrieval system using ontology includes a keyword mapping section to perform mapping of a keyword extracted from a retrieval query via ontology, and an ontology concept corresponding to this key word; a query graph determining section to determine the query graph from the ontology by using the mapped ontology concept; and an output node determining section to determine the output node, corresponding to the retrieval result of the retrieval query from the determined query graph. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an ontology-based content search system and method, and more particularly, to accurately grasp an intention of a search query of an incomplete user through the ontology, and provide content suitable for the query intention. The present invention relates to a content search system and method using an ontology.

  With the development of Internet search technology, providing accurate search results for user search queries has emerged as an important issue. The search ability varies depending on the user performing the Internet search, and the search queries that are actually input are diverse. Therefore, the search engine must accurately grasp the intention of the user's query and provide the search result desired by the user. is required.

  In particular, when information necessary for a search query at the time of search is missing, it is most important that the search engine grasps the intention of the user's search.

  However, since a search engine is a machine rather than a human, it is difficult to grasp the user's search intention from an incomplete user search query.

  For example, when the user inputs “USA, action”, it is assumed that the actual user's intention of searching is to search for a movie corresponding to the action genre among American movies. However, the search engine simply provides documents that contain both “America” and “Action” as a search result from documents contained in the database. There is a problem of providing a search result that has nothing to do with the user's search intention, such as a document with the content of "I took an action".

  Eventually, even if a search query that is incomplete and lacks information necessary for the search is input, a method for providing a search result that matches the user's search intention is required. That is, it is necessary to provide a configuration that gives a search engine, which is a machine, the ability to be inferred by using standardized human knowledge.

  As an embodiment of the present invention, a keyword constituting a user's search query is mapped to an ontology concept, and a query graph is determined on the ontology through the mapped ontology concept to complement an incomplete user's search query A content search system and method using an ontology are provided.

  In addition, as an embodiment of the present invention, there is provided a content search system and method capable of generating a query graph more quickly by mapping keywords and ontology concepts through a keyword-concept mapping table.

  Also, as an embodiment of the present invention, an unnecessary query graph is generated by generating a query graph in consideration of the maximum distance between nodes or whether data actually exists in the knowledge base. A content search system and method that can save resource consumption are provided.

  Furthermore, as one embodiment of the present invention, the query graph is determined by arranging the query graphs according to the degree of suitability with the search query among the generated query graphs, so that the intention of the user's search ascertained through the ontology is more accurate. The present invention provides a content search system and method using ontology that can be estimated.

  A content search system according to an embodiment of the present invention includes a keyword mapping unit that maps a keyword extracted from a search query via an ontology and an ontology concept corresponding to the keyword, and uses the mapped ontology concept. A query graph determination unit that determines a query graph from an ontology, and an output node determination unit that determines an output node corresponding to a search result of the search query from the determined query graph.

  According to an aspect of the present invention, the query graph determination unit generates at least one query graph including a node or an edge corresponding to the mapped ontology concept from the ontology, and aligns the generated query graph. A query graph suitable for the search query can be extracted from the at least one query graph.

  The content search system according to an embodiment of the present invention may further include a search result providing unit that provides a search result of a search query corresponding to the output node.

  The content search method according to an embodiment of the present invention includes a step of mapping a keyword extracted from a search query via an ontology and an ontology concept corresponding to the keyword, and the ontology using the mapped ontology concept. The method may include determining a query graph and determining an output node corresponding to a search result of the search query from the determined query graph.

  According to one aspect of the invention, determining a query graph from the ontology includes generating at least one query graph including nodes or edges corresponding to the mapped ontology concept from the ontology; Aligning the generated query graphs and extracting a query graph suitable for a search query from the at least one query graph.

  The content search method according to an embodiment of the present invention may further include providing a search result of a search query corresponding to the output node.

  According to an embodiment of the present invention, an incomplete user search query is performed by mapping the keywords that make up a user search query to an ontology concept and determining a query graph on the ontology via the mapped ontology concept. A content search system and method using an ontology is provided.

  In addition, according to an embodiment of the present invention, there is provided a content search system and method capable of generating a query graph more quickly by mapping keywords and ontology concepts via a keyword-concept mapping table. The

  In addition, according to an embodiment of the present invention, a query graph is considered in consideration of a condition regarding the maximum value of the distance between any two nodes of the query graph and / or whether data actually exists in the knowledge base. Providing a content search system and method that can save resources consumed to generate unnecessary query graphs.

  Furthermore, according to an embodiment of the present invention, the user's search intention ascertained through the ontology is determined by arranging the query graphs in accordance with the degree of matching with the search query among the generated query graphs. A content search system and method using ontology that can be estimated more accurately are provided.

FIG. 6 is a diagram for explaining a process of providing a search result for a search query using an ontology according to an embodiment of the present invention. 1 is a block diagram showing an overall configuration of a content search system according to an embodiment of the present invention. It is a figure for demonstrating the relationship between ontology and knowledge base which concern on one Embodiment of this invention. FIG. 5 is a diagram illustrating a process of mapping a keyword and an ontology concept through a keyword-concept mapping table according to an exemplary embodiment of the present invention. It is a figure which shows an example of the keyword-concept mapping table which concerns on one Embodiment of this invention. FIG. 6 is a diagram illustrating a combination of ontology concepts mapped to keywords via a keyword-concept mapping table according to an embodiment of the present invention. It is a figure which shows the query graph 1-1 determined from the combination 1 of ontology concept of FIG. 6 by one Embodiment of this invention. FIG. 7 is a diagram illustrating a query graph 1-2 determined from the ontology concept combination 1 of FIG. 6 according to an embodiment of the present invention. It is a figure which shows the query graph 2-1 determined from the combination 2 of ontology concept of FIG. 6 by one Embodiment of this invention. FIG. 7 illustrates a query graph 2-2 determined from ontology concept combination 2 of FIG. 6 according to one embodiment of the invention. It is a figure which shows the result of having substituted the query graph determined by one Embodiment of this invention by the data form. FIG. 6 is a diagram illustrating a process of determining an output node in a query graph through an output node determination rule according to an embodiment of the present invention. It is a figure which shows an example for demonstrating the output node determination rule by one Embodiment of this invention. 3 is a flowchart illustrating an entire process of a content search method according to an embodiment of the present invention.

  Hereinafter, embodiments according to the present invention will be described in detail with reference to the contents described in the accompanying drawings. However, the present invention is not limited or limited by the embodiments. The same reference numerals provided in each drawing denote the same members.

  FIG. 1 is a diagram illustrating a process of providing a search result for a search query using an ontology according to an embodiment of the present invention.

  Referring to FIG. 1, the user can input a search query 103. At this time, the search query 103 may include at least one keyword. The content search system according to an embodiment of the present invention can provide a search result 105 for a search query 103 input by a user. Here, the search query 103 may be in an incomplete form. In such a case, for example, when the search query 103 includes the keyword “A, B”, the intention of the user's search is to search for content including A and B at the same time, or A and B It is important to know whether it is to search for another content related to each other. That is, the user search query 103 may be configured with keywords from which important parts are omitted. In such a case, there is insufficient information for accurately grasping the user's search intention. There is a case.

  In this case, the content search system according to the embodiment of the present invention can complement the incomplete user search query 103 using the information model 100. That is, the content search system can provide a search result most suitable for the user's search intention by complementing the incomplete user search query 103 via the information model 100.

  As shown in FIG. 1, the information model 100 can be composed of an ontology 101 and a knowledge base 102. Ontology 101 may mean a schema area that represents class attributes and relationships between classes, and knowledge base 102 may mean a data area that actually represents data.

  As an example, ontology 101 can be classified into domain ontology such as global ontology, music ontology, movie ontology, and person ontology. Here, the domain ontology means an ontology 101 including domain knowledge suitable for a specific domain. At this time, the domain knowledge can correspond to the ontology concept nodes included in the domain ontology. Domain ontologies can be linked to other domain ontologies through concepts.

  The content search system can map the keyword extracted from the search query 103 to the ontology concept existing in the ontology 101 via the keyword-concept mapping table 104. Furthermore, the content search system can determine a query graph reflecting a user's search intention through an ontology concept and extract an output node from the query graph. Finally, the content search system can provide the search result 105 corresponding to the output node to the user.

  FIG. 2 is a block diagram showing the overall configuration of the content search system according to an embodiment of the present invention.

  Referring to FIG. 2, the content search system 200 may include a keyword mapping unit 201, a query graph determination unit 202, and an output node determination unit 203. In addition, the content search system 200 may include a search result providing unit 204.

  The keyword mapping unit 201 can map the keyword extracted from the search query 205 via the ontology 207 and the ontology concept corresponding to the keyword. Here, the ontology 207 may have a directional graph (directed graph) structure. Further, the ontology 207 may be composed of a plurality of nodes and edges for mapping with keywords extracted from the search query 205. At this time, each of the node and the edge existing in the ontology 207 may be defined by an ontology concept, and the edge may be an attribute of the node. In ontology 207 which is a logical model, a node-edge-node may be converted into one RDF triple (Resource Description Framework Triple).

  Furthermore, ontology 207 can be linked to knowledge base 208 that actually represents the data. The knowledge base 208 may also have a directional graph structure, and may be composed of knowledge base concepts including nodes and edges of the knowledge base 208. According to one embodiment of the present invention, nodes and edges of ontology 207 can be defined by ontology concept, and nodes and edges of knowledge base 208 can be defined by knowledge base concept.

  The graph structure having the directionality possessed by the ontology 207 and the graph structure having the directionality possessed by the knowledge base 208 can associate a predetermined area existing in the storage device of the computer with a node, for example. In this case, when the first area and the second area exist in the storage device of the computer, by storing the address of the second area in the first area, from the node associated with the first area, An edge to a node associated with the second region can be expressed.

  Alternatively, predetermined areas existing in the storage device of the computer can be associated with nodes and edges, respectively. In this case, the first node-edge-second node structure has an area associated with the edge, an address of the area associated with the first node, and an address of the area associated with the second node. It can be expressed by storing. In addition, the address of the region associated with the edge is stored in the region associated with the first node, and further, the address of the region associated with the second node is stored in the region associated with the edge. You may make it express by.

  The keyword mapping unit 201 can map keywords and ontology concepts using the keyword-concept mapping table 206. The keyword-concept mapping table 206 can be expressed by being stored in a storage device of a computer. As an example, the keyword mapping unit 201 can search the ontology concept corresponding to the keyword extracted from the search query 205 using the keyword-concept mapping table 206 and extract a combination of the ontology concepts that can be generated. At this time, the keyword-concept mapping table 206 may include a plurality of ontology concepts corresponding to one keyword.

  For example, when the keyword is “monster”, the ontology concept corresponding to the keyword may be “movie” or “alias”. As an example, the keyword-concept mapping table 206 may be an index data structure that maps and records keywords that are natural language expressions of the ontology 207 and the knowledge base 208 to the ontology concept. The keyword-concept mapping table 206 will be specifically described with reference to FIGS.

  The query graph determination unit 202 can determine a query graph from the ontology using the mapped ontology concept. As an example, the query graph determination unit 202 can generate at least one query graph including nodes or edges corresponding to the ontology concept mapped from the ontology 207. At this time, a plurality of query graphs may be generated for one keyword-ontology concept combination.

  The query graph can be expressed by using the storage area and address of the computer in the same manner as the graph structure having the directionality of the ontology 207 and the graph structure having the directionality of the knowledge base 208. Also, the query graph can be created by copying the structure of the subgraph including the ontology concept to which the keyword is mapped out of the graph structure of the ontology 207 to the storage device of the computer.

  As an example, the query graph determination unit 202 may be generated such that the maximum distance between nodes corresponding to the ontology concept is satisfied in a generated query graph. Here, the maximum distance between the first concept node and the second concept node is defined as the number of nodes existing on the path connecting the first concept node and the second concept node. Further, when there are a plurality of paths connecting the first concept node and the second concept node, the number of nodes existing on the path can be obtained for each path and defined as the maximum value. In addition, a weight can be set for the node, and the sum of the weights of the nodes existing on the path connecting the first concept node and the second concept node is calculated by the first concept node and the second concept node. It can also be the distance between. In this case, if the weights of all the nodes are set to 1, the sum of the weights matches the number of nodes. For example, if node 3 and node 4 exist on the path connecting node 1 and node 2, the weight of node 3 is 2, and if the weight of node 4 is 1, the distance between the nodes is 2 + 1 = 3 According to the embodiment of the present invention, the query graph determination unit 202 imposes a condition on the maximum value of the distance between any two nodes of the query graph, and the query graph having a large distance between nodes. By removing, it is possible to prevent a plurality of query graphs from being generated.

  As another example, the query graph determination unit 202 can generate a query graph using nodes and edges whose data actually exist in the knowledge base 208. In other words, it is not necessary to generate a query graph if there is no actual data for the nodes of a specific query graph in the knowledge base 208.

  Further, the query graph determination unit 202 can extract a query graph suitable for the search query 205 from at least one query graph by arranging the generated query graphs. At this time, the query graph determination unit 202 can extract the query graph suitable for the search query 205 by arranging the generated at least one query graph according to the degree of matching with the search query 205. Here, the query graph determination unit 202 can preferentially extract a query graph having a relatively short distance between a node and an edge corresponding to the ontology concept among at least one query graph. Alternatively, the fitness may be the size of the query graph. Here, the size of the query graph is defined using the maximum value of the distance between any two nodes of the query graph. For example, a query graph with a smaller graph size may have a higher matching degree.

  Through such a process, the content search system 200 can comprehend an incomplete user search query by using a query graph using the ontology concept, thereby grasping the intention of the user search.

  The output node determination unit 203 can determine an output node corresponding to the search query search result from the determined query graph. As an example, the output node determination unit 203 can determine the output node by repeatedly executing the route inspection according to the output node determination rule from the query graph.

  At this time, the output node determination unit 203 extracts a single group composed of a subject, a predicate, and an object from the query graph, and selects the object from the single group in advance. One node of the subject or the predicate can be determined depending on whether or not it is recognized. Furthermore, the output node determination unit 203 can determine a node as an output node in consideration of whether or not the determined node is a subject of another single group. Here, a single group can be defined as a triple on ontology.

  The search result providing unit 204 can provide the search result of the search query corresponding to the output node. For example, the search result providing unit 204 may provide a search result according to any one of CPC (Click Per Cost), CTR (Click Through Ratio), Quality Index (Quality Index), or a random method. Can be provided in line.

  FIG. 3 is a diagram for explaining a relationship between an ontology and a knowledge base according to an embodiment of the present invention.

  Referring to FIG. 3, it can be seen that the ontology 301 and the knowledge base 302 have a directional graph structure. The ontology 301 represents partial knowledge of the movie domain.

  In ontology 301, “Actor, Movie, Director” may represent a class. At this time, the class means to represent a set of instances having common attributes, and can correspond to one node in the ontology 301. “Actor” and “Movie” have a relationship of “hasMovie”. At this time, “hasMovie” can correspond to one edge in the ontology 301. In ontology 301, each node and edge can be defined by ontology concept.

  Further, “actor — 1273” of the knowledge base 302 connected to “Actor” of ontology 301 may mean an ID that identifies an actor with a unique name of Jeong Ji Hoon (or an actor with an alias of Pi). At this time, “actor — 1273” indicates an instance of the class “Actor”, which means actual data for the class. Similarly, “director_5738” of the knowledge base 302 connected to “Director” of the ontology 301 may mean an ID that identifies a director with a unique name, Park Chanok.

  “Actor — 1273” and “movie — 9842” can correspond to one node in the knowledge base 302, and “hasMovie” indicating the relationship between “actor — 1273” and “movie — 9842” can correspond to one edge. In the knowledge base 302, each node and edge can be defined by a knowledge base concept.

  The graph structure of the ontology 301 and the knowledge base 302 is a logical model, and may be stored after being converted into an RDF triple form when actually stored. At this time, the node-edge-node of the graph can be converted into one RDF triple.

  FIG. 4 is a diagram illustrating a process of mapping keywords and ontology concepts through a keyword-concept mapping table according to an embodiment of the present invention.

  Referring to FIG. 4, a keyword-concept mapping table 401, ontology 402, and knowledge base 403 are shown.

  When a search query 404 including at least one keyword is input from a user, the content search system can parse the search query 404 to extract keywords. The content search system can map the extracted keyword and the ontology concept of the ontology 402 to each other. As an example, the content search system can map keywords and ontology concepts using the keyword-concept mapping table 401.

  The keyword-concept mapping table 401 may be an index data structure recorded by mapping keywords that are natural language expressions of the ontology 402 and the knowledge base 403 to the ontology concept. Referring to the keyword-concept mapping table 401 shown in FIG. 4, the keyword “pi” can be mapped to the ontology concept “Actor-hasOtherName”. Also, the keyword “Im Soo-jung” can be mapped to the ontology concept “Actor-hasName”, and the keyword “movie” can be mapped to the ontology concept “O :: Movie”. Here, the keywords “Pi” and “Im Soo Jung” can be mapped to the ontology concept of the data type (class-attribute). Furthermore, since the keyword “movie” itself means a class, it can be mapped to the ontology concept of a class (Movie). However, O :: may be added in front of the class name in order to distinguish it from the data type.

  At this time, the graph structure shown in the ontology 402 is generated through a combination of keywords and mapped ontology concepts. Multiple graph structures may be derived for a combination of one ontology concept.

  FIG. 5 is a diagram showing an example of a keyword-concept mapping table according to an embodiment of the present invention.

  In the keyword-concept mapping table of FIG. 5, the keyword means ontology and knowledge-based natural language expression. Furthermore, the ontology concept can mean a node or attribute corresponding to a keyword in the ontology.

  In the case of ontology concept, natural language expressions can be set in a label form. For example, as shown in FIG. 3, the label “Movie” may be set for the “Movie” class, and the label “Actor” or “Movie Actor” may be set for the “Actor” class. In addition, a label of “appearing movie” or “appearing” may be set for the attribute “hasMovie” of the “Actor” class. The set label can be indexed from the keyword-concept mapping table.

  Furthermore, a plurality of ontology concepts may exist for a specific keyword. Referring to FIG. 5, in the ontology whose domain is a movie, the keyword “name” may be the name of a movie actor or the name of a movie director. Therefore, the keyword “name” can have the ontology concept “O :: Actor-hasName” or “O :: Director-hasName”.

  Through such a keyword-concept mapping table, keywords extracted from a search query can be more quickly mapped to ontology concepts on the ontology.

  FIG. 6 is a diagram illustrating a combination of ontology concepts that are mapped to keywords via a keyword-concept mapping table according to an embodiment of the present invention.

  Referring to FIG. 6, it is assumed that the search query 601 includes a keyword “Pi, Im Soo Jung, Movie”. The content search system can parse the search query 601 to extract at least one keyword. The content search system can parse the search query in consideration of the search query fragmentation. As an example, the content search system can parse the search query considering the keyword-concept mapping table 602.

  For example, if the search query is “Cyborg is OK”, the content search system cannot extract keywords separately from “Cyborg” and “OK” that do not exist in the keyword-concept mapping table. It can be extracted as a single keyword "But it's okay".

  When keywords that can be separated are combined to form a search query (for example, a Japanese movie), the content search system can parse and parse the keywords. When transliterating a foreign language, the content search system can extract keywords in consideration of various transliteration notations. Furthermore, the content search system can also construct a keyword-concept mapping table for synonyms, similar words, and related words for the extracted keywords.

  The keywords extracted in this way can be mapped to ontology concepts via the keyword-concept mapping table 602. FIG. 6 shows a concept mapping result 603 obtained by mapping the keyword extracted from the search query 601 to the ontology concept.

  If the keyword “pi” is an actor alias or a movie title, the keyword “pi” can be mapped to an ontology concept (Actor-hasOtherName) and (Movie-ofTitle). For this reason, the concept mapping result 603 can also be determined by being classified into the combination 1 and the combination 2, and thereafter, a query graph can be generated by the combination of the concept mapping result 603. The content search system can use the domain knowledge of the ontology to estimate a combination that is most relevant to the user's search intention among the mapped ontology concept combinations.

  7 to 10 show examples of query graphs that can be generated from the combination of ontology concepts shown in FIG. If the specific condition is set, the query graphs shown in FIGS. 7 to 10 may not be generated. For example, the specific condition may be a condition regarding the maximum distance between two nodes arbitrarily selected from the query graph. Further, it may be whether data actually exists in the knowledge base.

  The content search system can determine a query graph from the ontology using the mapped ontology concept. At this time, the content search system may generate at least one query graph including nodes or edges corresponding to the ontology concept mapped from the ontology.

  As an example, the content search system can generate a query graph in which the maximum distance between nodes corresponding to the ontology concept satisfies a predetermined condition. Examples of the predetermined condition include a condition that the maximum distance is not more than a predetermined value. The predetermined value is, for example, a value stored in a memory of the content search system so as to be readable, or a value given by a user at the time of search and temporarily stored in the memory. Furthermore, the content search system can generate a query graph in which data actually exists in the knowledge base.

  FIG. 7 is a diagram illustrating a query graph 1-1 determined from the ontology concept combination 1 of FIG. 6 according to an embodiment of the present invention.

  Referring to the results of FIG. 6, FIG. 7 shows that the ontology concept for the keywords “Pi”, “Im Soo Jung”, and “Movie” constituting the search query 701 is “ A process of generating a query graph 1-1 regarding the combination 1 mapped to “Actor-hasOtherName”, “Actor-hasName”, and “O :: Movie” is shown. That is, through the mapped ontology concept, Pi can mean the actor's alias, Im Soo-jung can mean the actor's name, and the movie can mean the movie itself.

  As shown in FIG. 7, it can be seen that a query graph 1-1 for the ontology concept combination 1 is generated using the ontology concept of the ontology 703. Through the query graph of FIG. 7, the intention of the user's search can be interpreted as searching for a movie in which an actor whose alias is “Pi” and whose name is Lim Soo-Jung has appeared.

  If there is no actor whose alias is Pi and whose name is Im Soo-jung, the query graph of FIG. 7 may be meaningless. Insignificant query graphs may eventually be eliminated when determining a query graph suitable for the user's search query.

  FIG. 8 is a diagram illustrating a query graph 1-2 determined from the ontology concept combination 1 of FIG. 6 according to an embodiment of the present invention.

  Referring to the result of FIG. 6, FIG. 8 shows that the ontology concept for the keywords “Pi”, “Im Soo Jung”, and “Movie” constituting the search query 801 is “ The process of generating a query graph 1-2 for the combination 1 mapped to “Actor-hasOtherName”, “Actor-hasName”, and “O :: Movie” is shown. That is, through the mapped ontology concept, Pi can mean the actor's alias, Im Soo-jung can mean the actor's name, and the movie can mean the movie itself.

  As shown in FIG. 8, it can be seen that the query graph 1-2 for the ontology concept combination 1 is generated using the ontology concept of the ontology 803. Through the query graph of FIG. 8, the intention of the user's search can be interpreted as searching for a movie in which an actor whose alias is Pi and an actor whose name is Im Soo-Jung have appeared.

  FIG. 9 is a diagram illustrating a query graph 2-1 determined from the ontology concept combination 2 of FIG. 6 according to an embodiment of the present invention.

  Referring to the results of FIG. 6, FIG. 9 shows that the ontology concept for the keywords “Pi”, “Im Soo Jung”, and “Movie” constituting the search query 901 is “ The process of generating a query graph 2-1 for the combination 2 mapped to “Movie-ofTitle”, “Actor-hasName”, and “O :: Movie” is shown. That is, through the mapped ontology concept, Pi can mean the title of a movie, Im Soo Jung can mean the name of an actor, and a movie can mean the movie itself.

  As shown in FIG. 9, it can be seen that the query graph 2-1 for the ontology concept combination 2 is generated using the ontology concept of the ontology 903. Through the query graph of FIG. 9, the intention of the user's search can be interpreted as searching for a movie whose title is “Pi” out of movies in which an actor whose name is Lim Soo Jung has appeared.

  FIG. 10 is a diagram illustrating a query graph 2-2 determined from the ontology concept combination 2 of FIG. 6 according to an embodiment of the present invention.

  Referring to the result of FIG. 6, FIG. 10 shows that the ontology concept for the keywords “Pi”, “Im Soo Jung”, and “Movie” constituting the search query 1001 is “ The process of generating a query graph 2-2 for the combination 2 mapped to “Movie-ofTitle”, “Actor-hasName”, and “O :: Movie” is shown. That is, through the mapped ontology concept, Pi can mean the title of a movie, Im Soo Jung can mean the name of an actor, and a movie can mean the movie itself.

  As shown in FIG. 10, it can be seen that the query graph 2-2 for the ontology concept combination 2 is generated using the ontology concept of the ontology 1003. Through the query graph of FIG. 10, the intention of the user's search can be interpreted as searching for a movie whose title is “Pi” among the movies in which an actor whose name is Lim Soo Jung appeared.

  FIG. 11 is a diagram illustrating a result of replacing a query graph determined by an embodiment of the present invention with a data format. At this time, FIG. 11 shows a query graph suitable for a user search query in FIGS. 7 to 10.

  The content search system can determine a query graph from the ontology using the mapped ontology concept. Specifically, the content search system can generate at least one query graph that includes nodes or edges that correspond to ontology concepts mapped from ontology.

  At this time, the content search system can generate a query graph in which the maximum distance between nodes corresponding to the ontology concept satisfies a predetermined condition. Here, the maximum distance between nodes may mean a value obtained by applying a sum of weights of nodes existing on the route in a route connecting the first concept node and the second concept node corresponding to the ontology concept. it can. Further, the content search system can generate a query graph in which data actually exists in the knowledge base.

  Furthermore, the content search system can extract a query graph suitable for a search query from at least one query graph by arranging the generated query graphs. At this time, the content search system can preferentially extract a query graph having a relatively small distance between a node corresponding to the ontology concept and an edge from at least one query graph.

  If there is no actor whose alias is Pi and whose name is Im Soo-jung, the query graph 1-1 shown in FIG. 7 is a meaningless graph and may not be generated.

  Also, if there is no case where an actor whose name is Im Soo-Jung appears in a movie titled Pi, the query graph 2-1 shown in FIG. 9 is also a meaningless graph, so it may not be generated. .

  Furthermore, when the maximum distance between nodes is set to 2, since the maximum distance of the query graph 2-2 shown in FIG. 10 is 3, the query graph 2-2 may not be generated from the beginning.

  Through such a process, the query graph 1-2 shown in FIG. 8 can be finally determined. If two query graphs are determined, the content search system can determine a query graph more suitable for the user's search intention.

  In order to determine the output node from the determined query graph, it is necessary to replace the data type node with the data node. Referring to FIG. 11, in the query graph 1-2, the content search system can replace a node having a hasOtherName attribute with respect to an Actor node with “pi”, and a node having a hasName attribute with respect to an Actor node. Can be replaced with “Im Soo Jung”. At this time, in the query graph 1-2, which node is determined as an output node may be a problem.

  FIG. 12 is a diagram illustrating a process of determining an output node in a query graph through an output node determination rule according to an embodiment of the present invention.

  The content search system can determine the output node by repeatedly performing the route inspection according to the output node determination rule from the determined query graph. Specifically, the content search system extracts a single group composed of a subject, a predicate, and an object from the query graph, and determines whether the subject or the subject is recognized in advance in the single group. Any one of the predicates can be determined. Furthermore, the content search system can determine a node as an output node in consideration of whether or not the determined node is the subject of another single group.

  Referring to FIG. 12, a query graph is input in step S1201. Further, in step S1202, the content search system extracts a single group (single SPO) from the query graph. Here, a single group means a subgraph of node (subject) → edge (predicate) → node (object) in the query graph.

  In step S1203, the content search system determines whether the object (Object: O) has already been recognized. Whether or not the object is already recognized can be determined by whether or not data actually exists (that is, whether or not the object is a data node).

  If the object is not recognized, the content system selects the object in step S1204. On the other hand, if the object is recognized, the content system selects the subject (Subject: S) in step S1205.

  In step S1206, the content search system determines whether the determined node is the subject of another single group. If the determined node becomes the subject of another single group, the content system returns to step S1202 and repeatedly executes the output node determination rule. On the other hand, if the determined node is not the subject of another single group, the content search system determines the determined node as an output node in step S1207.

  FIG. 13 is a diagram illustrating an example for explaining an output node determination rule according to an embodiment of the present invention.

  Referring to FIG. 13, the query graph 1-2 of FIG. 11 is shown. First, the content search system can extract a single group (SPO) from a query graph. As an example, the content search system can extract a single group from the terminal portion of the query graph.

In the first loop, the content search system is the first single SPO
Can be extracted. Is the first single group
, “Actor” means a subject, “hasOtherName” means a predicate (Property), and “pi” means an object (Object).

Similarly, in the first loop, the content search system is a second single SPO
Class can be extracted. Where is the second single group
, “Actor” means “subject”, “hasName” means “predicate”, and “im sujeong” means “object”.

  Thereafter, the content search system can determine whether or not the object is recognized in advance in a single group. In the above-described example, the object words “Pi” and “Im Soo-jung” are input as keywords and are data nodes. Therefore, the content search system can select the actor that is the subject in each of the single groups described above.

In the second loop, the content search system can determine whether the selected actor is the subject of another single group. Single group included in query graph 1-2
When
, It can be seen that Actor is the subject of another single group. That is, the actor that is the subject of the first single SPO is also the subject of a single group whose object is Movie, and the actor that is the subject of the second single SPO is a single subject whose subject is Movie. It is also the subject of the group. Thereafter, the content search system can repeatedly apply the output node determination rule.

Single group in the second loop
and
In this case, since the movie that is the object is not recognized in advance, the movie that is the object can be selected in a single group.

  Furthermore, since the Movie is not the subject of another single group, the loop is terminated, and finally the Movie node can be determined as the output node. After all, if the output node determined from the query graph 1-2 is taken into consideration, the content search system will search for a movie in which the user's search intention is an actor whose alias is Pi and an actor whose name is Im Soo Jung. It can be understood that the search has been made. Thereafter, the content search system can provide the user with a search result regarding a movie in which an actor whose alias is Pi and an actor whose name is Lim Soo-Jung appear.

  FIG. 14 is a flowchart showing an entire process of the content search method according to the embodiment of the present invention.

  In step S1401, the content search system parses the search query and extracts at least one keyword. As an example, the content search system can parse the search query taking into account the splitting and keyword-concept mapping tables.

  In step S1402, the content search system maps keywords and ontology concepts using the keyword-concept mapping table. The ontology concept can be composed of nodes and edges that are mapped to keywords in an ontology that is a graph structure with directionality. At this time, the edge may be an attribute of the node.

  Furthermore, the ontology can be linked to a knowledge base that actually represents the data. Like the ontology, the knowledge base has a directional graph structure, and can be constituted by a knowledge base concept including knowledge base nodes and edges.

  As an example, the content search system can search for ontology concepts corresponding to keywords using a keyword-concept mapping table, and extract combinations of ontology concepts that can be generated. At this time, there may be at least one ontology concept corresponding to the keyword. The keyword-concept mapping table may be an index data structure in which keywords, which are natural language expressions of ontology and knowledge base, are mapped and recorded with ontology concepts.

  In step S1403, the content search system generates at least one query graph from the ontology using the ontology concept. At this time, the content search system can generate at least one query graph including nodes or edges corresponding to the ontology concept from the ontology.

  As an example, the content search system can generate a query graph in which the maximum distance between nodes corresponding to the ontology concept satisfies a predetermined condition. At this time, the maximum distance between the nodes may be a value obtained by applying a weighted sum of the nodes existing on the route in the route connecting the first concept node and the second concept node corresponding to the ontology concept. .

  As another example, the content search system can generate a query graph in which data actually exists in the knowledge base.

  In step S1404, the content search system sorts the query graphs according to the degree of matching with the search query to determine a query graph suitable for the search query. As an example, the content search system can preferentially extract a query graph having a relatively short distance between a node and an edge corresponding to the ontology concept from at least one query graph. That is, the content search system can determine a query graph having a relatively small distance between a node and an edge from a plurality of query graphs as a query graph suitable for the search query.

  In step S1405, the content search system determines an output node corresponding to the search query search result from the query graph. As an example, the content search system can determine the output node from the determined query graph by repeatedly executing the route check shown in the flowchart of FIG. 12 according to the output node determination rule.

  Specifically, the content search system extracts a single group composed of a subject, a predicate, and an object from a query graph, and determines whether the subject or predicate is recognized in advance in the single group. Any one of the nodes is determined. Furthermore, the content search system determines whether the determined node is the subject of another single group, and determines the previously determined node as the output node.

  In step S1406, the content search system searches for and provides a search result of a search query corresponding to the output node. As an example, the content search system may provide the search results in an aligned manner according to any one or combination of CPC, CTR, quality index, or random method.

  For parts not described in FIG. 14, the description of FIGS. 1 to 13 can be referred to.

  The content search method using ontology according to an embodiment of the present invention includes a computer-readable recording medium including program instructions for executing various operations realized by a computer. The recording medium may include program instructions, data files, data structures, etc. alone or in combination, and the recording medium and program instructions may be specially designed and configured for the purposes of the present invention, It may be known and usable by those skilled in the computer software art. Examples of computer-readable recording media include magnetic media such as hard disks, flexible disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floppy disks, and ROMs. , Hardware devices specially configured to store and execute program instructions, such as RAM, flash memory, and the like. Further, the recording medium may be a transmission medium such as a light or metal line including a carrier wave for transmitting a signal that stores program instructions, a data structure, and the like, and a waveguide. Examples of program instructions include not only machine language code generated by a compiler but also high-level language code that can be executed by a computer using an interpreter or the like.

  As described above, the present invention has been described with reference to the preferred embodiments. However, those skilled in the relevant technical field should not depart from the spirit and scope of the present invention described in the claims. Thus, it will be understood that the present invention can be variously modified and changed. In other words, the technical scope of the present invention is defined based on the claims, and is not limited by the best mode for carrying out the invention.

100: Information model 101: Ontology 102: Knowledge base 103: Search query 104: Keyword-concept mapping table 105: Search result

Claims (33)

A keyword mapping unit that maps a keyword extracted from a search query via an ontology to an ontology concept corresponding to the keyword;
A query graph determination unit that determines a query graph from the ontology using the mapped ontology concept;
An output node determination unit that determines an output node corresponding to a search result of the search query from the determined query graph;
Content search system using ontology including The ontology concept is
A graph structure having a direction composed of nodes and edges mapped to the keywords via the ontology;
The edge is
The content search system according to claim 1, wherein the content search system is an attribute of the node. The ontology is
Linked to a knowledge base that actually represents data of the ontology,
The knowledge base is
The content search system according to claim 2, wherein the content search system has a graph structure having a direction composed of knowledge base concepts including nodes and edges of the knowledge base. The keyword mapping unit
The content search system according to claim 1, wherein the keyword is mapped to the ontology concept using a keyword-concept mapping table. The keyword mapping unit
5. The content search system according to claim 4, wherein the ontology concept corresponding to the extracted keyword is searched using the keyword-concept mapping table, and a combination of the ontology concepts that can be generated is extracted. The keyword-concept mapping table is:
5. The content search system according to claim 4, further comprising an index data structure recorded by mapping the ontology and a keyword that is a knowledge-based natural language expression to the ontology concept. The query graph determination unit
Generating at least one query graph including nodes or edges corresponding to the mapped ontology concept from the ontology;
The content search system according to claim 1, wherein the generated query graphs are arranged to extract a query graph more suitable for a search query from the at least one query graph. The query graph determination unit
The content search system according to claim 7, wherein a query graph is generated in which a maximum distance between nodes corresponding to the ontology concept satisfies a predetermined condition. The maximum distance between the nodes is
9. The value according to claim 8, wherein in a path connecting the first concept node and the second concept node corresponding to the ontology concept, the value is a sum of weights of nodes existing on the path. Content search system. The query graph determination unit
The content search system according to claim 7, wherein a query graph in which data actually exists in the knowledge base is generated. The query graph determination unit
The content search system according to claim 7, wherein the at least one query graph is sorted according to a degree of matching with the search query, and a query graph suitable for the search query is extracted. The query graph determination unit
The content search system according to claim 11, wherein a query graph having a relatively small distance between a node and an edge corresponding to the ontology concept is preferentially extracted from the at least one query graph. . The output node determination unit includes:
The content search system according to claim 1, wherein an output node is determined by repeatedly executing a route check according to an output node determination rule from the determined query graph. The output node determination unit includes:
A single group composed of a subject, a predicate, and an object is extracted from the query graph, and one of the subject and the predicate is determined depending on whether or not the object is previously recognized in the single group. 14. The content search system according to claim 13, wherein one node is determined. The output node determination unit includes:
The content search system according to claim 14, wherein the node is determined as an output node in consideration of whether or not the determined node is a subject of another single group.   The content search system according to claim 1, further comprising a search result providing unit that provides a search result of a search query corresponding to the output node. The search result providing unit
The content search system according to claim 16, wherein the search results are arranged and provided according to a criterion or a combination of any one of CPC, CTR, quality index, and random method. Mapping a keyword extracted from a search query via an ontology to an ontology concept corresponding to the keyword;
Determining a query graph from the ontology using the mapped ontology concept;
Determining an output node corresponding to a search result of the search query from the determined query graph;
Content search method using ontology including The ontology concept is
A graph structure having a direction composed of nodes and edges mapped to the keywords via the ontology;
The edge is
The content search method according to claim 18, wherein the content search method is an attribute of the node. The ontology is
Linked to a knowledge base that actually represents data of the ontology,
The knowledge base is
The content search method according to claim 19, wherein the content search method includes a graph structure having a directionality, which includes a knowledge base concept including nodes and edges of the knowledge base. Mapping the ontology concept comprises:
Searching the ontology concept corresponding to the keyword using the keyword-concept mapping table;
Extracting a combination of the ontology concepts that can be generated;
The content search method according to claim 18, comprising: The keyword-concept mapping table is:
The content search method according to claim 21, further comprising an index data structure recorded by mapping the ontology and a keyword that is a knowledge-based natural language expression to the ontology concept. Determining a query graph from the ontology includes:
Generating from the ontology at least one query graph including nodes or edges corresponding to the mapped ontology concept;
Aligning the generated query graphs and extracting a query graph more suitable for a search query from the at least one query graph;
The content search method according to claim 18, comprising: Generating the at least one query graph comprises:
The content search method according to claim 23, wherein a query graph is generated in which a maximum distance between nodes corresponding to the ontology concept satisfies a predetermined condition. The maximum distance between the nodes is
25. A value obtained by applying a sum of weights of nodes existing on the path on a path connecting the first concept node and the second concept node corresponding to the ontology concept. Content search method. Generating the at least one query graph comprises:
The content search method according to claim 23, wherein a query graph in which data actually exists in the knowledge base is generated. Extracting a query graph suitable for a search query from the at least one query graph,
The content search method according to claim 23, wherein the at least one query graph is sorted according to a degree of matching with the search query, and a query graph more suitable for the search query is extracted. Extracting a query graph suitable for a search query from the at least one query graph,
28. The content search method according to claim 27, wherein a query graph having a relatively small distance between a node and an edge corresponding to the ontology concept is preferentially determined from the at least one query graph. . Determining the output node comprises:
19. The content search method according to claim 18, wherein the output node is determined by repeatedly executing a route check according to an output node determination rule from the determined query graph. Determining the output node comprises:
Extracting a single group composed of subject, predicate and object from the query graph;
Determining any one node of the subject or predicate according to whether or not an object is previously recognized in the single group;
Selecting the node as an output node considering whether the determined node is the subject of another single group;
The content search method according to claim 29, comprising: Providing a search result of a search query corresponding to the output node;
The content search method according to claim 18, further comprising: Providing a search result of the search query comprises:
32. The content search method according to claim 31, wherein the search results are arranged and provided according to any one of CPC, CTR, quality index, or random method, or a combination thereof.   A computer-readable recording medium having recorded thereon a program for executing the method according to any one of claims 18 to 32.