Concept Embedding for Information Retrieval
This work addresses the term-mismatch problem in information retrieval, potentially improving conceptual indexing, but it appears incremental as it builds on existing word embedding methods.
The paper tackled the term-mismatch problem in information retrieval by developing three approaches to build concept vectors from word embeddings, enabling a vector-based similarity measure between concepts and showing promising experimental results.
Concepts are used to solve the term-mismatch problem. However, we need an effective similarity measure between concepts. Word embedding presents a promising solution. We present in this study three approaches to build concepts vectors based on words vectors. We use a vector-based measure to estimate inter-concepts similarity. Our experiments show promising results. Furthermore, words and concepts become comparable. This could be used to improve conceptual indexing process.