SOFTWARE DEVELOPMENT OF KEYWORD SEARCH AND PERFORMANCE ANALYSIS OF KEYWORD PERSONALIZATION ALGORITHM IN LARGE RDF GRAPH

• Main Author: Norman Saviero, Jehian
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/43780
• Institution: Institut Teknologi Bandung

Keyword search in a graph is a new research and start to be introduced at 2010. In this final report will be discussed about an approach to get (an) answer(s) from one or more keywords. This approach emphasizes finding the approximate resemblance of a graph using new metrics brought by Sinha and her colleagues. Still from the same researcher, to get candidate of the answer(s) in a large graph, will be used new graph that represent the large graph to speed up finding and to make easier to get simple form of candidate answer. After that, choose graph that more suitable to the user using approximation metrics from Sinha et al and take it to find the similar graph with the expected graph. According to analysis result, personalization keyword search has exponential complexity that causes the algorithm at the first would be not scalable. After doing next step of analysis, there is variable in which can be controlled so that the complexity of the algorithm can be reducted into polynomial. The variable is the number of keyword input. In the experiment phase, will be done using test case from the QALD because QALD already has gold standard. For each test case, will be find the answer of each test case separately and will be calculated the F1 score separately too. After that will be find the average/mean of F1 score that will be used to compare with SPARQL semantic matching algorithm. According to that result, keyword personalization algorithm still can’t beat the effectiveness level to answer the query compared to SPARQL semantic matching algorithm.