Home // eKNOW 2025, The Seventeenth International Conference on Information, Process, and Knowledge Management // View article

ColBERT-Based User Profiles for Personalized Information Retrieval

Authors:
Aleena Ahmad
Gibson Nkhata
Abdul Rafay Bajwa
Hannah Marsico
Bryan Le
Susan Gauch

Keywords: Personalized information retrieval, user profile generation, ColBERT, reranking algorithms.

Abstract:
Personalized Information Retrieval (PIR) improves search relevance by tailoring results to user interests using query history and browsing patterns. Traditional approaches to personalization range from feature engineering to the use of ontologies. Recently, there has been an increase in the exploration of deep learning models for this purpose. These models, such as Contextual Late Interaction over BERT (ColBERT), provide token-level contextual embeddings that can be leveraged to generate semantic user profiles. State-of-the-art approaches use ColBERT to select candidate terms for personalized query expansion from user profiles. This approach poses challenges in accurately choosing user’s descriptive keywords, risking the omission of crucial user preferences and repetitive selection of user terms. This study proposes a novel PIR approach that fully encodes user profiles using contextual embeddings and reranks Best Matching 25 (BM25) retrieved documents. Additionally, a frequency-recency weighting mechanism is tested which adjusts query influence based on temporal proximity and repetition frequency. Experimental results on two publicly available datasets demonstrate that our method improves retrieval performance, providing more accurate and context-aware search results.

Pages: 51 to 58

Copyright: Copyright (c) IARIA, 2025

Publication date: May 18, 2025

Published in: conference

ISSN: 2308-4375

ISBN: 978-1-68558-272-2

Location: Nice, France

Dates: from May 18, 2025 to May 22, 2025