Information retrieval with query expansion and re-ranking: a survey
Kaouthar Djoudi, Zaïa Alimazighi, Badiâa Dellal Hedjazi · IET conference proceedings. · 2025
The exponential data volume, variety, and velocity increase demands robust tools for rapid and relevant information access. This paper addresses the challenges of information retrieval (IR) systems, focusing on enhancing Query Expansion (QE) and passage re-ranking processes to improve relevance and efficiency. Traditionally, QE has improved the performance of IR models. Still, the advent of Artificial Intelligence (AI) techniques has led to the development of neural QE models that utilize embeddings to compute semantic similarities between queries and documents. Passage re-ranking, a machine learning task, estimates relevance scores between a query and candidate passages. Unlike traditional methods that rely on lexical similarities, modern approaches aim to capture semantic and contextual features beyond mere word matching. This paper first examines classical IR techniques and then explores contemporary methods, such as deep learning, with a particular emphasis on Transformers. We propose a semantic information search meta-model and conduct a comparative analysis, suggesting future directions for integrating semantic aspects to enhance QE and re-ranking.