Content-Based Filtering Recommendation in Abstract Search Using Neo4j

Abstract

© 2017 IEEE. In this work, we focus on development of a content search on report documents and recommendation on related document from search result. The main contribution of this work is to model document content into graph. Document-Keyword graph was created to represent the relationship between document and its features. The data were stored as a connected graph in Ne04j graph database. The graph were used to filter keyword co-occurrence documents in order to reduce search space. The performance of the proposed model was evaluated with accuracy 0.77. To improve the accuracy, the model can be extended with collecting user selection as collaborative feedback to the system, or extended with domain specific ontology to analyze the semantic relationship of the documents.