KR100993845B1 - System For Recommending Personalized Meaning-Based Web-Document And Its Method - Google Patents

Abstract

The present invention relates to a personalized semantic based web document recommendation system and method. The present invention comprises: (a) collecting at least one web document using a web robot; (b) Each of the collected at least one web document is defined as an instance, and an identifier capable of identifying the corresponding web document is defined as an instance name of the instance, so that the generalized ontology of the at least one web document is collected. Generating; (c) monitoring a web document visited by the specific user; (d) the system inferring a concept (hereinafter referred to as “user preference concept”) included in the web document visited by the specific user from the generalized ontology using an ontology inference engine; (e) the system generating at least one instance of at least one concept most similar to the user preference concept among the concepts included in the generalization ontology; (f) the system receiving a web document recommendation request from the specific user; And (g) when the system receives the document recommendation request, recommending at least one web document from the generalization ontology to the specific user based on the user preference concept, first similarity, and second similarity. Provides a method for recommending personalized semantic based web documents.

delete

Description

System for Recommending Personalized Meaning-Based Web-Document And Its Method}

The present invention relates to a web document recommendation field, and more specifically, to generate a generalization ontology defining a semantic relationship from documents collected by a web robot in order to sort out a document requested by a user, and to generate a generalization ontology from a document visited by a user. Personalized semantic-based web document that creates a preference information ontology that reflects the user's preference concept and preference by inferring association with concepts, and recommends a web document that matches the user's intention using generalization ontology and preference information ontology A recommendation system and method.

In the field of web search, users spend a lot of time to find desired information. Various web document recommendation systems have been developed for efficient web search.

The web document recommendation system according to the related art is a method of recommending a web document preferred by a user through a similarity measurement using a frequency or a probabilistic method of a word based on a text mining technique, and the content of which the document is semantically related. Despite the implication, if the same word does not appear in the document, there is a problem that is not recommended.

In particular, a web document recommendation system that does not consider semantic relations is only capable of cross-sectional recommendation. For example, when a user is interested in "baseball," only documents containing the word "baseball" may be recommended. If semantic relations are considered, information on semantic related concepts such as "catcher" and "pitcher" associated with "baseball" can be recommended.

In addition, as research on the semantic web has recently been actively conducted, a semantic web-based search engine called a semantic portal has been developed, but there is a difficulty in providing connectivity with ontologies expressing documents and conceptual relationships. This causes a problem in which the semantic information required by the user in the web document recommendation field is accurately identified and results in a conflict with the desire to receive various information through conceptual association.

The present invention solves the above problems and is proposed according to recent trends and requests, and generates a generalization ontology defining semantic relationships from documents collected by a web robot to screen documents requested by a user, and the user Inferring the association with the concepts of generalized ontology from the visited documents, creating preference information ontology that reflects the user's preferred concept and preferences, and recommending a web document that matches the user's intention using the generalized ontology and preference information ontology The purpose is to provide a personalized semantic based web document recommendation system and method.

As an aspect of the present invention for achieving the above object, the personalized semantic-based web document recommendation method according to the present invention is a system for connecting a specific user through a network and recommending a web document to the specific user through the network It is implemented through, (a) collecting at least one web document by the system using a web robot; (b) the system defines each instance of each of the at least one web document collected, and defines an identifier capable of identifying the web document as an instance name of the instance, thereby relating to the collected at least one web document. Creating a generalized ontology; (c) the system monitoring the web document visited by the specific user; (d) the system inferring a concept (hereinafter referred to as “user preference concept”) included in the web document visited by the specific user from the generalized ontology using an ontology inference engine; (e) the system generating at least one instance of at least one concept most similar to the user preference concept among the concepts included in the generalization ontology; (f) the system receiving a web document recommendation request from the specific user; And (g) when the system receives the document recommendation request, recommending at least one web document from the generalization ontology to the specific user based on the user preference concept, first similarity, and second similarity. .
The step (d) indexes all instances of the generalization ontology into a vector model, and the user preference concept is the concept most similar to the concept included in the web document visited by the specific user from all the instances of the indexed generalization ontology. It consists of reasoning,
The step (f) is made by recognizing that the system receives a query term that the specific user queries the system, or selects a specific web document visited by the specific user,
The first similarity is the similarity (S1) between the query word and the instances of the generalized ontology, or the similarity (S2) between the specific web document and the instances of the generalized ontology, and the second similarity is the concept to which the query language belongs and the generalized ontology. Characterized in that the similarity (S3) between the concepts to which the instances belong or the concept included in the specific web document and the similarities (S4) between the concepts to which the instances of the generalization ontology belong.

The instance may include information regarding at least one of a comment, an image, a title, and a URL describing the corresponding web document.

The step (e) further includes generating preference information indicating preferences of the specific user for the created instance, wherein the preference of the user is the user's preference for the web document associated with the created instance. It may be characterized by being determined in relation to the frequency of visits.

The S1 and S2 may be calculated by calculating the cosine similarity between the query word and the instances of the generalized ontology, and the cosine similarity between the specific web document and the instances of the generalized ontology, respectively.

The S3 and S4 correspond to the intersection of the two sets in the first set consisting of at least one concept deduced from the query term or the specific web document and the concepts belonging to instances of the generalization ontology. It can be characterized in that it is calculated as how many concepts compared to the concepts corresponding to the union of the two sets.

delete

delete

delete

As another aspect of the present invention for achieving the above object, the personalized semantic-based web document recommendation system according to the present invention collects at least one web document using a web robot and collects the at least one web document. A web document collection unit that generates a generalization ontology for web documents; Monitor the web document visited by the user, infer a concept related to the web document visited by the user (hereinafter referred to as "user preference concept") from the generalization ontology, and associate the deduced user preference concept with the generalization ontology A preference information generating unit that generates an ontology of preference information about the user to which the user is assigned; And a document recommendation unit recommending at least one web document from the generalization ontology to the user in consideration of the user preference concept and document similarity according to a document recommendation request received from the user.

According to the personalized meaning-based web document recommendation system and method according to the present invention, a semantic relationship is defined from a document collected by a web robot to generate a generalization ontology, and a user's preference concept and preference are grasped from a document visited by the user By creating a preference information ontology, the generalized ontology and the preference information ontology can be used to recommend a personalized web document accurately reflecting the user's intention.

The above-mentioned objects, features and advantages of the present invention will become more apparent through the following detailed description in conjunction with the accompanying drawings. Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. Throughout the specification, the same reference numbers refer to the same components. In addition, when it is determined that a detailed description of a known function or configuration related to the present invention may unnecessarily obscure the subject matter of the present invention, the detailed description will be omitted.

1 is a block diagram of a personalized meaning-based web document recommendation system according to an embodiment of the present invention. 2 is a flowchart of a method for recommending a personalized semantic based web document according to an embodiment of the present invention. 1 and 2 and the necessary drawings, a personalized semantic based web document recommendation system and method according to an embodiment of the present invention will be described in detail.

The personalized meaning-based web document recommendation system 100 may include a web document collection unit 20, a preference information generation unit 40, and a document recommendation unit 50, as shown in FIG. 1. . The functions of the web document collection unit 20, the preference information generation unit 40, and the document recommendation unit 50 are briefly described as follows.

The web document collection unit 20 collects at least one web document using a web robot and generates a generalization ontology 24 for the collected at least one web document.

The preference information generation unit 40 infers the concept included in the web document visited by the logged-in user from the generalization ontology 24, and uses the inferred concept as the user preference concept of the user, so that the preference for the user The information ontology 45 is created.

The document recommendation unit 50 recommends at least one web document to the user using the general ontology 24 and the preference information ontology 45 according to a document recommendation request received from the user.

Hereinafter, a detailed operation of the personalized meaning-based web document recommendation system 100 and a method of recommending a personalized meaning-based web document according to an embodiment of the present invention will be described in detail.

The web document collection unit 20 collects at least one web document using a web robot [S101]. The web robot may collect all documents connected by a hyperlink to the starting point of the initially given Seed URL 10. At this time, the web document collection unit 20 may use a queue (Queue, 21). Q is a queue. The web document collection unit 20 stores the first given Seed URL 10 in the queue 21 and collects documents corresponding to the Seed URL 10 stored in the queue 21 [S101]. All URLs included in the document corresponding to the Seed URL 10 are obtained [S103] and added to the queue 21 [S105]. Then, the web document collection unit 20 may collect documents corresponding to URLs stored in the queue 21 [S101]. In this way, the web document collection unit 20 may repeatedly perform steps S101 to S105 a predetermined number of times.

The web document collection unit 20 generates a generalization ontology 24 for the collected at least one web document [S107]. The generalization ontology 24 grasps the subject matter of the collected at least one web document and defines conceptual connectivity.

3 is a diagram illustrating an example of generating a generalized ontology. In general, an ontology about a web document is composed of information related to concepts, instances, and associations between concepts included in the web document. The web document collection unit 20 may generate the generalized ontology 24 in a manner that is defined as each different instance for each of the at least one web document collected. In addition, each of the instances defines an identifier 60 capable of identifying the web document as an instance name, and an annotation 61, image 62, and title 63 describing the web document. ) And URL 64.

The preference information generating unit 40 monitors the web document 41 visited by a user who is logged in to the system 100 (hereinafter referred to simply as a'user' as a specific user who has requested a web document recommendation) [S109] ], the concept included in the web document 41 visited by the user (hereinafter referred to as "user preference concept") is inferred from the generalized ontology 24 using the ontology inference engine 46 [S111]. Here, the preference information generating unit 40 indexes all instances of the generalization ontology 24 into a vector model, and the concept most similar to the concept included in the web document 41 visited by the user is the user preference concept. Can deduce.

4 is a view showing an example of the step S111. The step S111 is a step of inferring which of the concepts belonging to the generalization ontology 24 is a web document 41 visited by the user in order to predict what kind of document the user is interested in. to be. As illustrated in FIG. 4, the preference information generation unit 40 indexes annotations 61 of all instances included in the generalization ontology 24 into a vector model 70 and the ontology reasoning engine 46 ) To target the instances of the generalized ontology 24 indexed using the concept most similar to the concept contained in the web document 41 visited by the user as the concept included in the web document 41 visited by the user Can deduce. Referring to FIG. 4, inference of the concept by step S111 may be performed using the Concepts function 71, which receives the query q and measures the cosine similarity with the document d indexed in the vector model. And returns the set of concepts included in the document with the highest similarity.

A vector model is a model that expresses a document on a vector to grasp the association between documents. To express a document on a vector, all words appearing in the document can be expressed as TF (Term Frequence)-IDF (Inverse Document Frequence). For example, documents A and B may be expressed in five dimensions when "a b c c e" is expressed in the document A and document "a b c c e" is expressed in the document B. If only TF is considered, document A is (a, b, c, d, e)=(1, 1, 2, 1, 0), and document B is (a, b, c, d, e)=(1, 1, 2, 0, 1). IDF lowers the weight of words that appear too frequently considering the frequency of documents. For example, the word a appears in both document A and document B, so it is not a good word to distinguish between documents. Therefore, the weight of the word a is low, and the weight of the word d or e is high. In this way, all documents are represented on a vector by an index.

A cosine measurement method may be used to measure similarity between documents expressed in the vector model. The cosine measurement method is a method of normalizing two documents as a unit vector and measuring a cosine angle between the two documents, as indicated by reference numeral 71 in FIG. 4. The cosine measurement method is a method commonly used in vector models of information retrieval.

In addition, the preference information generation unit 40 generates a preference information ontology 45 for the user, which has been given a correlation between the inferred user preference concept and the generalized ontology 24 [S113]. 5 is a diagram showing an example of generating the preference information ontology 45.

In step S113, at least one concept most similar to the user preference concept among the concepts included in the generalization ontology 24 is generated as at least one instance of the preference information ontology 45, and the at least one most similar. The concept may be associated with the at least one instance. For example, in step S113, as illustrated in FIG. 5, among the concepts included in the generalization ontology 24, at least one concept 80 most similar to the user preference concept is the preference information ontology 45 ), and the association between the most similar at least one concept and the instance 81 by using the "referTo" ObjectProperty.

5 shows the relationship between the generalized ontology 24 and the preference information ontology 45. For example, when the user is interested in the concept of "Machine Learning" of the generalization ontology 24, as shown in FIG. 5, the "Machine Learning" instance 81 in the preference information ontology 45 Is created and a "referTo" relationship is formed as the source from which this concept is preferred. In the example of FIG. 5, the background in which the user prefers the concept of “Machine Learning” is that a “referTo” relationship is formed because a document suitable for an instance called “Bayesian Network” has been previously visited.

In step S113, preference information indicating the user's preference for the generated instance 81 may be generated. The preference of the user indicates how much the user prefers the created instance 81. The preference of the user may be determined in relation to the frequency of the user's visit to the web document associated with the created instance 81. Referring to FIG. 5, the user's preference may be expressed by having a “hasWeight” relationship. That is, in the preference information ontology 45, the concepts preferred by the user are expressed as instances, and through the "referTo" relationship, it is possible to know which similar web document was previously visited, and through the "hasWeight" relationship, the user You will know how much you prefer a particular concept or specific instance.

6 is a diagram illustrating an example of a method of measuring a user's preference for the generated instance 81. As illustrated in FIG. 6, the preference information generation unit 40 uses the instances of the preference information ontology 45 as a key, and the weight value of the “hasWeight” DatatypeProperty associated with the instance corresponding to the key. The user's preferences can be measured or modified using the PreferenceDic and the deduced concepts of the document the user is currently visiting. The dictionary includes user preference information for each concept.

The preference information generation unit 40 determines whether all the concepts deduced from the document visited by the user are included in the generated dictionary key set (PreferenceDic.KeySet). If not all of the deduced concepts are included, the value of the concept (c) is initialized to 0 for only those not included in the dictionary (NewConcepts) (PreferenceDic[c]=0). The initializing process is for specifying that the previous preference of the concepts is 0. The weight for the currently visited document is determined by the "Occur" function. If the concept of the document currently visited by the user matches the concept of PreferenceDic, a value of 0 may be returned.

The preference information generation unit 40 may update the preferences in consideration of the previous preferences and whether or not the concept of the document visited by the user 83 for all concept keys of the added dictionary (82). For example, when a user first visits a web document of the “network” concept, PreferenceDic[network]=1 because it is a concept not defined in the preference ontology 45. Then, when the user first visits a web document of the "database" concept, PreferenceDic[network]=0.5 and PreferenceDic[database]=1 (σ=0.5). The degree to which the value of the previous preference becomes smaller by the σ value may be set. When the user visits "network" again, PreferenceDic[network]=1.26, PreferenceDic[database]=0.5. Therefore, the preference for documents that are not frequently visited gradually decreases to a value close to zero. As a result, a high preference may be generated for a concept that the user is recently interested in or a concept that is frequently interested.

The document recommendation unit 50 receives a document recommendation request from the user [S115]. The document recommendation request may be received by receiving a query term inquired by the user or recognizing that the user selects a web document. For example, it is recognized that a web document is selected by the user, for example, selecting a specific web document that the user visits.

The document recommendation unit 50 may infer similar concepts and documents of the specific query word or the specific web document from the preference information ontology 45 using the ontology reasoning engine 46 [S117]. The ontology inference engine 46 may perform inference of up-down relationships of concepts and new concept inference of instances. Examples of the ontology inference engine 46 include FACT++, Pellet, Kaon2, RacerPro, and the like.

For example, when the concept of “upper” is defined as “the person who throws the ball” and “the baseball”, the ontology reasoning engine 46 is the sub-concept of the person who throws the ball. Infer that it is a sub-concept of baseball.

As an example of new concept inference of an instance, when there is an "baseball" concept instance, "Lee Seung-yeop", the ontology reasoning engine 46 infers a conceptual relationship so that "Lee Seung-yeop" is a sub-concept of the "baseball" concept. Infer that it is an instance of "other".

The document recommendation unit 50 uses the ontology inference engine 46 to infer the up-down relations of concepts inferred from the generalization ontology 24 and based on the changed up-down relations, a document similar to a query entered by the user. The concept can be inferred as the concept of a query.

The document recommendation unit 50 recommends at least one web document from the generalization ontology 24 to the user using the user preference concept and document similarity according to the document recommendation request [S119]. The document recommendation unit 50 may recommend a document suitable for the user's intention by considering the similarity of the vector model, the weight of the user preference concept, and the concept weight of the user input query.

7 is a view showing an example of a formula for performing step S119. As illustrated in FIG. 7, the document recommendation unit 50 targets instances of the generalized ontology 24 indexed, the similarity of the vector model (85), and the similarity between the concept of the query and the concept included in the document. (86), at least one web document may be recommended by calculating the score 84 considering the weight 82 of the user preference concept and sorting in descending or ascending order.

The similarity 85 of the vector model means the similarity (S1) between the specific query language and the instances of the generalized ontology 24 or the similarity (S2) between the specific web document and the instances of the generalized ontology 24. The similarity 85 of the vector model (S1 and S2 are respectively) calculates the cosine similarity between the specific query language and the instances of the generalized ontology 24, and the cosine similarity between the specific web document and the instances of the generalized ontology 24. Can be calculated. That is, the similarity 85 of the vector model may be calculated by calculating the cosine similarity between the query word input by the user and the indexed document in the indexed vector model.

The similarity 86 between the concept to which the specific query term belongs and the concept included in the document is the similarity between the concept to which the specific query word belongs and the concepts to which the instances of the generalization ontology 24 belong (S3) or the specific web document includes It means the similarity (S4) between the concept and concepts to which the instances of the generalized ontology 24 belong. The similarity 86 between the concept to which the query word belongs and the concept included in the document (S3 and S4) is a first set consisting of at least one concept deduced from the specific query word or the specific web document and the generalization ontology 24 ) In a concept set including a second set of concepts to which instances of) belong, how many concepts corresponding to the intersection of the first set and the second set are compared to the union of the first set and the second set have. For example, the first set is inferred as {'baseball','catcher'}, and when the second set is {'baseball','pitcher'}, it includes the first set and the second set The concept set (union) is {'baseball','catcher','pitcher'}. Therefore, the similarity 86 between the concepts to which the query language belongs and the concepts to which the instances of the generalized ontology 24 belong is 1/3. In addition, for example, if the first set is {'baseball','catcher'} and the second set is also {'baseball','catcher'}, the concept to which the query language belongs and the instance of the generalization ontology 24 The similarity 86 between the concepts to which they belong is 2/2. That is, the similarity 86 between the concept to which the query language belongs and the concept included in the document indicates how much these two concepts overlap.

8 is a view showing an example of a document recommended by step S119. "Network" was input as a recommended keyword in the interface 90 that collects documents visited by the user and provides recommendation information. The concept of the recommended document is shown on the right side of the screen 92, and documents are listed on the left side of the screen using priority. When a user requests a document recommendation regarding "network", as shown in FIG. 8, the concept preferred by the user so far can be given a higher weight to recommend a personalized semantic-based document.

The personalized semantic based web document recommendation method according to the present invention described above may be provided by recording on a computer-readable recording medium as a program for executing on a computer.

The personalized semantic based web document recommendation method according to the present invention can be executed through software. When executed in software, the configuration means of the present invention are code segments that perform necessary tasks. The program or code segments can be stored on a processor readable medium or transmitted by a computer data signal combined with a carrier wave in a transmission medium or communication network.

The computer-readable recording medium includes all types of recording devices in which data readable by a computer system is stored. Examples of computer-readable recording devices include ROM, RAM, CD-ROM, DVD±ROM, DVD-RAM, magnetic tape, floppy disk, hard disk, and optical data storage device. The computer-readable recording medium can also be distributed over network-connected computer devices so that the computer-readable code is stored and executed in a distributed fashion.

The present invention described above, as the person skilled in the art to which the present invention pertains, various substitutions, modifications and changes are possible without departing from the technical spirit of the present invention. It is not limited by the drawings.

1 is a block diagram of a personalized meaning-based web document recommendation system according to an embodiment of the present invention.

2 is a flowchart of a method for recommending a personalized semantic based web document according to an embodiment of the present invention.

3 is a diagram illustrating an example of generating a generalized ontology.

4 is a view showing an example of the step S111.

5 is a diagram showing an example of generating the preference information ontology 45.

6 is a diagram illustrating an example of a method of measuring a user's preference for the generated instance 81.

7 is a view showing an example of a formula for performing step S119.

8 is a view showing an example of a document recommended by step S119.

<Explanation of reference numerals for main parts of drawings>

100: Web document recommendation system 10: Seed URL

20: Web document collection unit 21: Queue

24: generalization ontology 40: preference information generating unit

45: preference information ontology 46: inference engine

50: Document recommendation department

Claims (21)

A method of recommending a web document to a specific user using a system connected to a specific user through a network and recommending a web document to the specific user through the network: (a) the system collecting at least one web document using a web robot; (b) the system defines each instance of each of the at least one web document collected, and defines an identifier capable of identifying the web document as an instance name of the instance, thereby relating to the collected at least one web document. Creating a generalized ontology; (c) the system monitoring the web document visited by the specific user; (d) the system inferring a concept (hereinafter referred to as “user preference concept”) included in the web document visited by the specific user from the generalized ontology using an ontology inference engine; (e) the system generating at least one instance of at least one concept most similar to the user preference concept among the concepts included in the generalization ontology; (f) the system receiving a web document recommendation request from the specific user; And (g) when the system receives the document recommendation request, recommending at least one web document from the generalization ontology to the specific user based on the user preference concept, first similarity, and second similarity, The step (d) indexes all instances of the generalization ontology into a vector model, and the user preference concept is the concept most similar to the concept included in the web document visited by the specific user from all the instances of the indexed generalization ontology. It consists of reasoning, The step (f) is made by recognizing that the system receives a query word that the specific user queries the system, or selects a specific web document visited by the specific user, The first similarity is the similarity (S1) between the query word and the instances of the generalized ontology, or the similarity (S2) between the specific web document and the instances of the generalized ontology, and the second similarity is the concept to which the query language belongs and the generalized ontology. A method of recommending a personalized semantic-based web document, characterized in that the similarity (S3) between concepts belonging to instances of S or the concept included in the specific web document and similarity (S4) between concepts belonging to instances of the generalization ontology. delete The method of claim 1, wherein the instance, A personalized semantic based web document recommendation method comprising information on at least one of a comment, an image, a title, and a URL describing the web document. delete delete The method of claim 1, wherein the step (e), The method further includes generating preference information indicating preferences of the specific user for the generated instance, The preference of the user, the personalized meaning-based web document recommendation method characterized in that it is determined in relation to the frequency of the user's visit to the web document associated with the created instance. delete delete The method of claim 1, wherein S1 and S2 are each A method for recommending a personalized semantic-based web document, wherein the cosine similarity between the query word and the instances of the generalized ontology is calculated by calculating the cosine similarity between the specific web document and the instances of the generalized ontology. The method of claim 1, wherein S3 and S4 are In the first set consisting of at least one concept deduced for the query word or for the specific web document and the second set consisting of concepts to which instances of the generalization ontology belong, concepts corresponding to the intersection of the two sets are the two A method for recommending personalized semantic-based web documents, characterized in that it is calculated as compared to the concepts corresponding to the set union. A computer-readable recording medium recording a computer program capable of executing the method of any one of claims 1, 3, 6, 9 or 10 with a computer. At least one web document is collected using a web robot, and each of the collected at least one web document is defined as an instance, and an identifier capable of identifying the web document is defined as an instance name of the instance. A web document collection unit generating a generalization ontology for the at least one web document; Monitor the web document visited by a specific user, infer the concept (hereinafter referred to as "user preference concept") included in the web document visited by the specific user from the generalization ontology using the ontology inference engine, and to the generalization ontology. A preference information generation unit generating at least one concept that is most similar to the user preference concept among the included concepts as at least one instance; And Receiving a document recommendation request from the specific user, and recommending at least one web document from the generalization ontology to the specific user based on the user preference concept, first similarity, and second similarity according to the received document recommendation request Include a document recommendation section, The preference information generating unit indexes all instances of the generalization ontology into a vector model, and the concept most similar to the concept included in the web document visited by the specific user from all the instances of the indexed generalization ontology as the user preference concept Reasoning, Receiving the document recommendation request is made by recognizing that the specific user receives a query term queried or selects a specific web document visited by the specific user, The first similarity is the similarity (S1) between the query word and the instances of the generalized ontology, or the similarity (S2) between the specific web document and the instances of the generalized ontology, and the second similarity is the concept to which the query language belongs and the generalized ontology. Personalized semantic-based web document recommendation system, characterized in that the similarity between the concepts to which the instances belong (S3) or the concepts included in the specific web document and the concepts to which the instances of the generalization ontology belong (S4). delete The method of claim 12, wherein the instance, Personalized semantic based web document recommendation system, characterized in that it comprises information about at least one of the annotation, image, title and URL describing the web document. delete delete The method of claim 12, wherein the preference information generating unit, Preference information indicating the user's preference for the generated instance is generated, The personal preference based web document recommendation system, characterized in that the user's preference is determined in relation to the frequency of the user's visit to the web document associated with the created instance. delete delete The method of claim 12, wherein S1 and S2 are each A personalized semantic-based web document recommendation system, characterized by calculating cosine similarity between the query word and the instances of the generalized ontology, and cosine similarity between the specific web document and the instances of the generalized ontology. The method of claim 12, wherein S3 and S4 are In the first set consisting of at least one concept deduced for the query language or for the specific web document and the second set consisting of concepts to which instances of the generalization ontology belong, concepts corresponding to the intersection of the two sets are the two Personalized semantic-based web document recommendation system, characterized in that it is calculated by how much compared to the concepts corresponding to the set union.

Images