KR102075833B1 - Curation method and system for recommending of art contents - Google Patents

Abstract

The present invention relates to an artwork recommendation curation system and method, the artwork recommendation curation system comprising: a user consumption history data storage unit for storing consumption history data of a user; A content metadata storage unit for storing the meta data of the artwork; Randomly show artworks, collect user preference information and purchase history information of the artworks, store them in the user consumption history data storage, and extract keywords from the descriptions of the artists created when the artworks are registered. A data collection unit for storing metadata about the artwork selected using the keyword in the content metadata storage unit; A data analysis unit analyzing a similarity between the user consumption history data collected by the data collection unit, a similarity between the artwork metadata, and a similarity between the user and the artwork by applying a statistical model based on gender, weather, and season; An indexing unit for indexing the data analyzed by the data analyzing unit to reduce a data size; And applying the data of the index to an art specialty ontology built on the basis of genre, emotion, image pattern, material, and color, and assigning a weight to the weather, season, and user's preference for the artwork. It includes a recommendation unit for generating personalized data.

Description

Curation method and system for recommending of art contents}

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to curation, and more particularly, to a method and system for recommending a curation of art by applying an ontology of art.

Recently, as the amount of content on the web has increased exponentially, a lot of time and effort has been required to select the content desired by users. In recent years, more and more consumers are buying by believing the information presented by experts. In the IT field, researches on 'recommended algorithms' and 'recommended systems' that can effectively select content are being actively conducted. 'Recommended system' recommends items that are of interest to the user, reducing the time for the user to select the content and at the same time promoting content consumption. Representative recommendation algorithms used in the recommendation system include collaborative filtering, cluster models, and search-based methods. The best examples of such algorithms are Amazon.com and Netflix. Currently, a content recommendation system is being developed in the fields of movies, books, and music.

However, the development of recommendation algorithm for art contents is not made. The art market includes artists as art producers, and consumers as individual collectors, companies, and institutions. The distribution market that connects these producers and consumers is divided into galleries, auction companies, and art fairs. In the representative off-line art market, transactions are centered on expensive works of famous artists. As famous artists are actively sold and resold, their prices continue to rise. However, new artists and unknown artists need to open the possibility that the works of these unknown artists can be sold because the entry into the art market is difficult. As an alternative, it is necessary to build an online art market that is accessible to everyone without cost constraints and location constraints, rather than a costly offline market.

If an efficient online art market is established, artists will be able to promote and sell their works, and consumers will be able to easily and friendlyly access artworks and purchase them at reasonable prices. In particular, it is possible to increase the credibility of the art work through transparent transactions, and to collect the works and to make a database, it is expected that the systematic work database management will be possible.

Publication No. 10-2017-0030379 (2017.03.17)

The problem to be solved by the present invention was created to solve the above problems of the art world, the buyers of artworks can own the work at a low price, and at the same time to help the artists from the work management to the preparation of the exhibition In order to achieve this, a model-based filtering recommendation engine that mixes content-based and user-based is provided to provide a method and system for recommending curation of artworks that apply art ontology.

Art work recommendation curation system according to the present invention for achieving the above technical problem, the user consumption history data storage unit for storing the consumption history data of the user; A content metadata storage unit for storing the meta data of the artwork; Randomly show artworks, collect user preference information and purchase history information of the artworks, store them in the user consumption history data storage, and extract keywords from the descriptions of the artists created when the artworks are registered. A data collection unit for storing metadata about the artwork selected using the keyword in the content metadata storage unit; A data analysis unit analyzing a similarity between the user consumption history data collected by the data collection unit, a similarity between the artwork metadata, and a similarity between the user and the artwork by applying a statistical model based on gender, weather, and season; An indexing unit for indexing the data analyzed by the data analyzing unit to reduce a data size; And applying the data of the index to an art specialty ontology built on the basis of genre, emotion, image pattern, material, and color and assigning a weight to weather, season, and user's preference for the artwork. It includes an artwork recommendation unit that generates personalized data.

The art work recommendation unit analyzes and calculates the similarity between contents, the similarity between users, and the similarity between users and contents using an item profile analyzing the artwork itself and a user profile extracted from the user's preferences and history. It is characterized by recommending a matching artwork to the user.

The art work recommendation curation system according to the present invention applies a topic association ontology to a user's query, finds a keyword related to the subject, measures similarity among contents, and groups high content of similarity to find and search for a representative keyword for each group. It further includes a semantic search unit.

According to an aspect of the present invention, there is provided a method for recommending a curation of a work of art, comprising: randomly displaying a work of art and collecting user preference information and purchase history information of the work of art; Extracting keywords from the descriptions of the artists created when the artworks are registered and collecting metadata about the selected artworks using the keywords; Analyzing the similarity between the collected user consumption history data, the similarity between the artwork metadata, and the similarity between the user and the artwork by applying a statistical model based on gender, weather, and season; Indexing the analyzed data to reduce the data size; And applying the indexed data to an art specialty ontology built on the basis of genre, emotion, image pattern, material, and color, and weighting the user's preference for the art and the weather and season. Generating personalized data for each user and recommending the same to the user.

The recommending step may be performed by analyzing and calculating similarity between contents, similarity between users, and similarity between users and contents using an item profile analyzing the artwork itself and a user profile extracted from user preferences and history. It is characterized by recommending a matching artwork to the user.

Art recommendation curation method according to the present invention comprises the steps of interpreting the user's query; Finding a topic association keyword by applying a topic association ontology to the interpreted user's query; Measuring the similarity between the contents and grouping the high similarity contents to find a representative keyword for each group; And searching by using the subject-related keyword and the representative keyword of each group.

According to the art work recommendation curation method and system according to the present invention, a user can easily and easily access a work of art that suits his or her taste, and the user can save a lot of time and effort in selecting the desired content.

According to the present invention, it is possible to provide a curation system applying an algorithm of an art content classification system suitable for an online service, and to provide an efficient social commerce platform for enjoying and selling art.

In addition, according to the present invention, the artists can promote and sell their works, and consumers can easily and friendlyly access the artworks and purchase the works at reasonable prices. In particular, transparent transactions are possible, so that the credibility of the artwork can be increased. In addition, the works are collected and made into a database, so that a systematic work database can be managed.

In addition, according to the present invention, it is possible to build a 'social curation commerce platform' utilizing the artwork recommendation algorithm and recommendation system that the general public can easily find and enjoy works of art that suit their tastes.

Figure 1 is a block diagram showing the configuration of an embodiment of the artwork recommendation curation system according to the present invention.
2 illustrates an example of recommending art content through a personalized recommendation engine reflecting consumer tastes.
3 is a schematic diagram illustrating an example of a curation system combining an item profile and a user profile.
4 is a flowchart showing an embodiment of a method for recommending a curation of art according to the present invention.
5 is a flowchart illustrating a semantic search process according to the present invention.
6 is a block diagram illustrating a concept of content classification processing.
7 illustrates an example of a triple structure between topic related keywords.
8 is a block diagram illustrating a concept of content analysis and classification processing.
9 shows an example of a conceptual diagram of generating and processing recommended content.
10 illustrates a first metadata processing process based on art content classification system data.
11 illustrates the SNS platform interworking process as a final API method after the completion of the secondary metadata processing.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. Configurations shown in the embodiments and drawings described herein are only one preferred embodiment of the present invention, and do not represent all of the technical spirit of the present invention, various equivalents which may be substituted for them at the time of the present application It should be understood that there may be variations and variations.

Figure 1 is a block diagram showing the configuration of an embodiment of the artwork recommendation curation system according to the present invention. Art recommendation curation system according to an embodiment of the present invention is a user consumption history data storage unit 110, content metadata storage unit 120, data collection unit 130, data analysis unit 140, indexing unit 150 and the artwork recommendation unit 160, and may further include a semantic search unit 170.

 The user consumption history data storage unit 110 stores the consumption history data of the user, the content metadata storage unit 120 stores the metadata of the artwork.

The data collecting unit 130 randomly shows the artwork and collects the user's preference information and the user's purchase history information on the artwork and stores it in the user consumption history data storage 110 and registers the artwork. The keyword is extracted from the descriptions of the artists, and the metadata for the selected artwork is stored in the content metadata storage 120 using the keywords.

The data analysis unit 140 applies a statistical model based on gender, weather, and season, and the similarity between the user consumption history data collected by the data collection unit 130 and the similarity between the artwork metadata and the user and the artwork. Analyze the similarity between the livers.

The indexing unit 150 reduces the data size by indexing the data analyzed by the data analyzing unit 140.

The artwork recommendation unit 160 applies the data of the index unit 150 to an art specialty ontology built on the basis of genre, emotion, image pattern, material, and color. By weighting the user's preferences to create a user-specific personalized data matching.

The artwork recommendation unit 160 analyzes and calculates the similarity between the contents, the similarity between the users, and the similarity between the users and the contents by using an item profile analyzing the artwork itself and a user profile extracted from the user's preference and history. We recommend the artworks that are matched through the ontology to the user.

The semantic search unit 170 finds a subject relation keyword by applying a subject relation ontology to a user's query, measures similarity among contents, and groups and finds a representative keyword for each group by grouping similar contents.

The art curation system according to the present invention easily registers the content (artwork) by the artist (producer) in the SNS and automatically gathers the content, classifying, indexing, and recommending the collected art content according to an efficient independent classification system, It can be used in the art-based social commerce platform, where consumers can easily select and purchase works.

In the consumer service scenario in the art specialized social commerce platform, when the consumer checks the information through a personalized recommendation engine, searches for the information of interest, and inputs his or her own information, a list of recommended works suitable for the user's taste is displayed on the screen. The consumer selects the work, checks the price of the work, the artist, the detailed image, etc., puts the selected work in the shopping cart, and finally purchases the work.

The producer uploads the artwork on a social commerce platform specializing in art. The upload is made by the artist painter selects an image of the image to be sold in the camera roll, inputs and uploads details of the artwork to be sold, and the uploaded artwork is automatically stored in my profile.

The artwork recommendation curation system according to an embodiment of the present invention may include a semantic search engine and a personalized recommendation engine.

The art recommendation curation system according to the present invention utilizes the characteristics of algorithms that can effectively select art contents, and applies art classification systems (ontologies) such as genres, emotions, image patterns, materials, and colors. I recommend artwork.

The recommendation algorithm of the artwork recommendation curation system is based on 'collaboration filtering based on contents'. Based on the algorithm for consumer purchase history, this is provided as an HTTP-based API. While the conventional collaborative filtering uses only user behavior records, the present invention enhances the similarity by using an item profile analyzing the content itself and a user profile extracted from the user's preference and history.

The semantic search process based on the item profile provides an integrated search using the art content classification system, and builds a semantic search system by forming an item profile of the academic and emotional classification systems related to the art content.

The personalized recommendation engine reflecting the consumer's tastes analyzes user profiles such as consumer preferences, appreciation, and purchase history, and recommends art content using a real-time API method. 2 illustrates an example of recommending art content through a personalized recommendation engine reflecting consumer tastes. Referring to FIG. 2, the user 210 inputs a rating into the rating input target content 220. The rating information may include a user's preference, a review rating, and a purchase rating. The rating input target content 220 is provided by the curation system 230 as a rating input target in the curation platform, and the rating input by the user 210 is transmitted to the curation system 230. Curation system 230 receives the rating information of the art content and stores in the rating DB. The real time profiling module generates a user profile with reference to the rating DB and outputs the real time curation result to the real time curation screen 240. The real time profiling module may generate an existing curation by training a training engine that receives the purchase log information.

The real-time curation screen 240 represents a result of combining the existing curation result and the received rating, and is reflected in the preferred genre and content curation in real time whenever the rating is input. The user 210 receives a curation in which the user's rating input is reflected in real time through the real time curation screen 240.

3 is a schematic diagram illustrating an example of a curation system combining an item profile and a user profile. Referring to FIG. 3, when the data collector 330 collects user consumption history data 310 and content metadata 320 data necessary for curation of an artwork, the analyzer 340 uses a statistical model and a user model. Analyze the similarity. The indexer 350 reads, writes, and schedules the index on the analyzed data in order to reduce engine overload and processing time, thereby creating an index file. The recommender 360 recommends the artwork using the data generated from the indexer, and transmits the artwork to the open API server 370. The open API server 370 recommends the artwork to the user.

In the art content classification system research (analysis technique), the art work recommendation curation system according to the present invention is conventionally used for archiving focused on values as academic data, while the present invention has advanced the curation system and has been optimized for popularization. The system is applied and classified according to user purpose, ie emotion or environment.

In the past, the present invention provided a content recommendation related technology using a user profile, a search engine related technology, and a user query analysis. However, the present invention provides a recommendation engine specialized for art content through a semantic search engine and a personalized recommendation engine related to art content. To improve the extraction of key words, headwords, and related keyword information when processing metadata, improve the recognition rate based on the meta-classification of art content, and to provide machine learning specialized to art content. The classification system of the existing classified art content and the classification system are classified as learning data that can be recognized together. In addition, it analyzes the user's query and builds information based on the emotional vocabulary so that the query can be interpreted based on the emotional recognition between the art content and the user.

4 is a flowchart showing an embodiment of a method for recommending a curation of art according to the present invention. 1 and 4, an embodiment of a method for recommending a curation of art according to the present invention will be described. First, the artwork is randomly displayed, and the user's preference information and the user's purchase history information are collected for the artwork (step S410). The keywords are extracted from the descriptions of the artists created when the artwork is registered and the keywords are extracted. Collect metadata about the selected artwork by using the step S420.

The data analyzer 140 analyzes the similarity between the collected user consumption history data, the similarity between the artwork metadata, and the similarity between the user and the artwork by applying a statistical model based on gender, weather, and season. S430 step)

The indexing unit 150 reduces the data size by indexing the analyzed data for reducing the engine overload and processing time (step S440).

The artwork recommendation unit 160 applies the indexed data to a professional art ontology built on the basis of genre, emotion, image pattern, material, and color, and weather, season, and user's preference for the artwork. By assigning a weight to each user to create a personalized data for each matching user is recommended to the user (step S450).

The recommendation step S450 may be performed by analyzing and calculating the similarity between contents, the similarity between users, and the similarity between the user and the content using an item profile analyzing the artwork itself and a user profile extracted from the user's preference and history. Through matching artwork can be recommended to the user.

5 is a flowchart illustrating a semantic search process according to the present invention. Referring to FIG. 5, a query of a user is analyzed (step S510). The subject-related keyword is found by applying a topic association ontology to the analyzed user query. (Step S520) The similarity between contents is measured and the similarity is measured. The high content is grouped together to find a representative keyword for each group (step S530). The search is performed using the subject-related keyword and the representative keyword for each group (step S540).

On the other hand, the present invention builds a work of art classification system (ontology) for recommending a work of art. To this end, the ontology designed as a work expression model is constructed based on the attributes of things, facts, and information. And for the semantic retrieval, we construct the topic association ontology based on the topic-based classification system and the topic relevance.

In order to organize case and experience-oriented ontology information as a classification system according to users, classification management is performed for classification of case input through classification map and classification criteria management.

The art content ontology is constructed to provide a curation system for art content recommendation. In order to provide an art content curation system, metadata is collected and processed, and metadata history is managed. In addition, it constructs and implements an art taxonomy model, classifies and maps artworks by extracting text mining and keywords and headings, generating and extracting dictionary information, and establishing classification association and similarity rules. The art curation system according to the present invention provides a query interpretation engine, an index engine, an integrated content search engine, and a customized recommendation engine. The query analysis engine analyzes user queries, extracts weights for each query, constructs rules related to pattern search, and analyzes morphemes. The index engine processes large volume of index data and has a data structure for index compression and fast search.

The integrated content search engine processes a large distributed index, has a fast search algorithm, and sorts and filters the search results. It also provides API functions, processes search logs, and processes search statistics data.

The custom recommendation engine generates personalized recommendation data for each individual and generates content-related recommendation data for personalized training, has a content-based related recommendation training engine, provides a recommendation API function, and sorts and filters for recommendations. It handles large amounts of index-based data.

The present invention can be used in an art social commerce platform. To this end, content that can increase user's interest can be exposed on the main screen, and various classification systems can be easily registered through the artwork registration screen. Provides page UX related to content registration, search and recommendation, and updates interest information when following and joining interested users and users through social networks. In addition, the main page personalization recommendation engine and API are linked.

In addition, for semantic retrieval, a topic-based ontology based on a topic-based classification system and topic relevance is built and supported.

The present invention can perform text mining. Text mining can find information that is meaningful and meaningful in metadata text data, and can find more than just information search, such as identifying the categories of text by linking with other information. In order to deeply analyze the information described in human language and to extract the information contained in it, a large amount of language resources and complex statistical algorithms must be applied. The technical areas that can be achieved through text mining are classification, clustering, information extraction, and summarization. The present invention implements a description of the 'Information Extraction' part based on the result of converting the voice query content of the customer to the text. By applying it to cognitive-based knowledge classification, the accuracy and satisfaction of knowledge recommendation can be improved.

In the present invention, the information extraction (Information Extraction) to analyze the information having a significant meaning in the text document to automatically extract the meta-property information, it can be used to extract a variety of information, such as keywords, key concepts, situations and conditions.

The function of automatic content (cognitive based) classification by text mining is to classify documents through machine learning based on statistical methods, to create rules to improve the accuracy of document classification, and to use statistical methods and flexible methods by rules. Combination can improve the accuracy of document classification. It can also provide a professional and convenient user interface for editing categories.

6 is a block diagram illustrating a concept of content classification processing. Referring to FIG. 6, content similarity and association are determined 1030 through content classification filtering 620, and statistical information processing 640 and content classification system mapping 650 are performed.

The construction of the theme association ontology can be performed in three stages. Extract the keywords using the stemmer for all the existing information (step 1). Find keywords that are related to each of the extracted keywords. (Step 2) Each of the extracted keywords and the corresponding keywords found in step 2. Each keyword associated with is ontology to fit the subject classification system in 'topic relevance' relation (step 3).

The related keyword extraction algorithm may be performed by searching for related content (Step 1), extracting related keyword candidates from related content (Step 2), and selecting related keywords (Step 3). The related content search (Step 1) searches for N social media content associated with a given keyword. Extracting related keyword candidates from related content (Step 2) collects keywords that are close to a given keyword in the related content message, creates a co-occurrence matrix of given keywords and proximity keywords, and generates DF (Document) for all proximity keywords. After generating a frequency vector and generating a TF (Term Frequency) vector for all neighboring keywords, a higher correlation keyword candidate is extracted by calculating a harmonic mean of the TF and DF vectors. Selecting the relevant keyword (Step 3) normalizes the frequency of the related keyword candidates by calculating similarities. The similarity includes string based similarity, SoundHex based similarity, Synonyms, same words in Korean and English. In addition, selecting the associated keyword (Step 3) removes stopwords and selects the top K keywords among the related keyword candidates sorted in descending order and sorted by the frequency value of the normalized related keyword candidates, thereby defining the related keyword.

The ontology of the related keywords may be stored in a triple structure between the subject related keywords as shown in FIG. 7.

In order to efficiently search for knowledge, the present invention provides a content indexing function and utilizes a topic association ontology for rich search results. It supports semantic search to find them.

The integrated search function has integrated search function of knowledge and derives search results through search term expansion. It uses synonym dictionaries to expand synonyms and synonyms for search terms, and uses the related relevance function and topic relevance ontologies for search terms. Supports integrated search to extend to include. Similar content search and representative keywords are extracted, and similarity between contents is measured to group contents with high similarity and representative keywords for each group are presented.

8 is a block diagram illustrating a concept of content analysis and classification processing. Referring to FIG. 8, the data manager 840 receives the content 810, the legacy DB 820, and the user input content 830, and analyzes the morpheme. Using a taxonomy model, the dictionary and text mining and association keywords are extracted. (850) Keyword-based indexing and retrieval (870) is performed through an associated content search engine with a structured metadata storage structure (860). To perform.

The indexing process extracts the text through content preprocessing, analyzes the morphemes on the extracted text, and indexes and stores the extracted text and the analysis results through the morphological analysis. The integrated search procedure inputs a search word from the user, analyzes the morpheme for the input search word, expands the search word through the analyzed result, and searches from the index using the extended search word. The search results performed are ranked according to the criteria and the integrated search results are provided.

Personalized recommendation engines analyze and train associations, preferences, statistics, logs, and dispositions between users and content to improve the ability to recommend appropriate knowledge of user requirements and respond to situations. The basics of recommendation are search engine indexing and compression techniques that take into account satisfaction with recommendation results, processing capacity for large amounts of data, and speed of data analysis. Content recommendation features include statistical information extraction function, user usage log analysis function, user propensity analysis function, user preference analysis function, and user-user association analysis function. Analysis of association between content and content, training and generation of personalized (curation) information for each user, training and creation and curation of related content information for each content, index of related recommendation information, user, content recommendation search result UI and curation engine There is an SNS platform.

9 shows an example of a conceptual diagram of generating and processing recommended content. Referring to FIG. 9, the correlation analysis and content details, content classification system, various statistical information, user logs, user usage statistics, content tags, etc. are received, and existing training information, preference information, training system setting information, recommendation result usage statistics, etc. The recommendation information is generated by referring to the content information and the training information excluding the recommendation. The recommendation information includes association recommendation and personalization.

Meanwhile, the SNS platform interworking process will be described in more detail based on the first metadata processing process, the second metadata processing method, and the API method based on the art content classification system.

10 illustrates a first metadata processing process based on art content classification system data. Referring to FIG. 10, meta data is data that can be classified as unlimited data and add additional information continuously. When data is recorded, an index means attributes such as the name, data size, etc. of the data and its recording location. It means to display such as a table. First, at the time of registering a work, raw data is collected through a description of a work 1010 and the query interpretation engine 1020 interprets the query. Analyze user query, extract weights for each query, search patterns, build related rules, and implement stemming analysis technology. After analyzing the query, metadata is collected 1030 and metadata processing and metadata history are processed 1040. The content is automatically classified through the art classification system based on basic cognition (1050). Map the classification scheme (1060). At this time, it recommends automatic classification and automatic classification system based on cognition, and extracts keywords suitable for the art classification system through the query interpretation engine from the description of the work when registering the works through machine learning technology and incorporates them into the classification system to autonomously upgrade the classification system. . After mapping the classification system, it processes the large index data, compresses the index, structures the data for high-speed search, and processes the large data without overload through the index linkage agent (1070). The search results are sorted and filtered through an engine that can efficiently search for stored content. (1080)

11 illustrates the SNS platform interworking process as a final API after completion of secondary metadata processing. Referring to FIG. 11, content-oriented metadata is processed through a content-based profile including content segments and content history information. (1110) Content statistics based on gender, age, region, weather, season, etc. Do the processing. Then, the similarity between the contents is analyzed to generate recommendation information. (1120) Through this, a group of similar contents is formed.

Meanwhile, the user-centric metadata is processed through a user-based profile including user segments and user history information (1130). For example, content statistics processing is performed based on gender, age, region, weather, season, and the like. Then, the similarity between users is analyzed to generate recommendation information. (1140) Through this, a group of similar users is formed.

Data is stored through an index engine (1150), and data is searched through a content-integrated search engine (1160). Also, personalized recommendation engines generate customized data for each user, and similarity and content-related recommendation data are generated between users. (1170) Then, the result is implemented by using the recommended API method and output. (1180)

The present invention can be embodied as code that can be read by a computer (including all devices having an information processing function) on a computer-readable recording medium. Computer-readable recording media include all kinds of recording devices that store data that can be read by a computer system. Examples of computer-readable recording devices include ROM, RAM, CD-ROM, magnetic tape, floppy disks, optical data storage devices, and the like. In addition, in this specification, “unit” may be a hardware component such as a processor or a circuit, and / or a software component executed by a hardware component such as a processor.

Although the present invention has been described with reference to the embodiments shown in the drawings, this is merely exemplary, and it will be understood by those skilled in the art that various modifications and equivalent other embodiments are possible. Therefore, the true technical protection scope of the present invention will be defined by the technical spirit of the appended claims.

110: user consumption history data storage
120: content metadata storage unit
130: data collection unit 140: data analysis unit
150: index unit 160: artwork recommendation unit
170: semantic search unit 210: user
220: Rating input content 230: Curation system
240: real-time curation screen 310: user consumption history data
320: content metadata 330: data collector
340: Analyzer 350: Indexer
360: recommender 370: open API server

Claims (6)

A user consumption history data storage for storing consumption history data of the user;
A content metadata storage unit for storing the meta data of the artwork;
Randomly show artworks, collect user preference information and purchase history information of the artworks, store them in the user consumption history data storage, and extract keywords from the descriptions of the artists created when the artworks are registered. A data collection unit for storing metadata about the artwork selected using the keyword in the content metadata storage unit;
A data analysis unit analyzing a similarity between the user consumption history data collected by the data collection unit, a similarity between the artwork metadata, and a similarity between the user and the artwork by applying a statistical model based on gender, weather, and season;
An indexing unit for indexing the data analyzed by the data analyzing unit to reduce a data size;
The data of the index is applied to an art ontology built on the basis of genre, emotion, image pattern, material, and color, and weighted on the weather, season, and user's preference for the artwork. A recommendation unit generating custom data; And
Find the subject-related keywords by applying the subject-related taxonomy based on the subject-based taxonomy and the subject relevance to the user's query, measure the similarity between the contents, and group the highly similar contents to find the representative keywords by group. A semantic search unit for searching the consumption history data storage unit and the content metadata storage unit by using a related keyword and the representative keyword of each group, and outputting a search result for a user query;
The recommendation part
Analyze the similarity between contents, similarity between users, and similarity between users and contents by using the item profile analyzing the artwork itself and the user profile that reflects user's preferences, reviews, and purchase scores, and extracts the user's purchase history and consumption history. Recommend the artwork to the user through the art professional ontology by calculating,
The subject-related ontology
Extract the keywords using the stemming analyzer for the artwork information, find the keywords related to the extracted keywords (associated keywords), and ontology to fit the subject classification system with the extracted keywords and the related keywords in the subject relation. The related keyword is selected by extracting a related keyword candidate from related content, and the related keyword candidate extraction collects keywords (proximated keywords) in close proximity to a given keyword in the related content message and extracts DF ( A recommendation curation system for art, characterized in that for generating a Document Frequency (TF) vector and a TF (Term Fequency) vector, and calculating the harmonic mean of the TF and DF to extract an associated keyword candidate. delete delete Randomly displaying the artwork and collecting user's consumption history data including user's preference information and purchase history of the artwork;
Extracting keywords from the descriptions of the artists created when the artworks are registered and collecting metadata about the selected artworks using the keywords;
Analyzing the similarity between the collected user consumption history data, the similarity between the metadata for the artwork, and the similarity between the user and the artwork by applying a statistical model based on gender, weather, and season;
Indexing the analyzed data to reduce the data size;
The indexed data is applied to an art specialty ontology constructed based on genre, emotion, image pattern, material, and color, and weighted to the weather, season, and user's preference for the artwork. Generating personalized personalized data and recommending the user to the user; And
Find the subject-related keywords by applying the subject-related taxonomy based on the subject-based taxonomy and the subject relevance to the user's query, and measure the similarity among the contents and group the contents with high similarity to find the representative keywords by group. Outputting a search result for a user query by searching the consumption history data of the user and metadata about the artwork by using the related keyword and the representative keyword of each group,
The recommendation step
By analyzing and calculating the similarity between the contents, the similarity between the users, and the similarity between the users and the contents by using the item profile analyzing the artwork itself and the user profile extracted from the user's preferences and consumption history, the artworks matched by the art specialty ontology Recommend to your users,
The subject-related ontology
Extract the keywords using the stemming analyzer for the artwork information, find the keywords (associated keywords) related to the extracted keywords, and use the ontology to fit the subject classification system in the subject relations with the extracted keywords and the related keywords. The related keyword is selected by extracting a related keyword candidate from related content, and the related keyword candidate extraction collects keywords (proximated keywords) in close proximity to a given keyword in the related content message and extracts DF ( And generating an associated keyword candidate by calculating a harmonic mean of the TF and the DF after generating a Document Frequency (TF) vector and a Term Fequency (TF) vector.
delete delete

Images