Keyword Search System Based On Klustree Method

摘要

KlusTree is a method for searching keywords into graph data. This method utilizes the use of language model to cluster keyword search results. The aim of this experiment is to build keyword search system using KlusTree method. The system accepts CSV files for data source and return keyword search results in a meaningful sentence form. Relational semantic mapping is used for converting CSV files into data with graph structure. TextRank and Depth First Search methods are used for modifying results in tree form to sentence form as cluster representation. The experiment shows that graph data produced by relational semantic mapping has directed graph structure. The quality of annotation on graph is affected by annotation in CSV files. The sentence form creates different perspective of keyword search result but it has a possibility to return ambiguous sentence.