Exploring Combinations of Ontological Features and Keywords for Text Retrieval

TL;DR

This paper investigates combining ontological features with keywords in vector space models to improve text retrieval, demonstrating better performance than traditional keyword-based methods like Lucene.

Contribution

It introduces novel adaptations of the Vector Space Model that integrate entity names, classes, and identifiers, enhancing retrieval effectiveness.

Findings

Proposed models outperform Lucene in retrieval tasks

Models improve document and query representation

Enhanced retrieval accuracy with ontological features

Abstract

Named entities have been considered and combined with keywords to enhance information retrieval performance. However, there is not yet a formal and complete model that takes into account entity names, classes, and identifiers together. Our work explores various adaptations of the traditional Vector Space Model that combine different ontological features with keywords, and in different ways. It shows better performance of the proposed models as compared to the keyword-based Lucene, and their advantages for both text retrieval and representation of documents and queries.