Paper 42 (Research track)

Graph Similarity Based Personalized Keyword Search on Large RDF Graphs

Author(s): Souvik Brata Sinha, Dimitri Theodoratos

Abstract: Keyword search is a popular technique to retrieve information from linked data on the web. With the increase in RDF graph data repositories, it becomes essential to work on a keyword search method that identifies user intent in order to rank the relevant results from a pool of huge number of candidate results for a keyword query. Our work is focused on personalized keyword search; where we intend to take into account some information about a user who is searching for a query, so that we can provide the results most relevant to the user. We extract the structural summary of the RDF graph and compute the pattern
graphs for a keyword query on the RDF graph using its structural summary. The user information is in the form of a graph similar to the pattern graphs called profile graphs. The ranking of the relevant results is achieved by measuring the graph similarity between the user profile graph and the generated pattern graphs. We propose similarity metrics which produce a similarity score to individual pattern graphs based on which the pattern graphs are ranked such that a pattern graph with a higher similarity score is ranked to be more relevant to the user. Our experimental results show that our approach intuitively matches the the user interests.

Keywords: Keyword Search; RDF Graph; Structural Summary; Personalization

Leave a Reply

Your email address will not be published. Required fields are marked *