D-MiQ: Deep Multimodal Interactive Healthcare Query Expansion Approach for Web Search Engines Retrieval Effectiveness

Health Information Retrieval (HIR) has become the new trend that tries to respond in an efficient, relevant, and satisfying way to specific users’ information requirements. The dilemma is that the health care domain ought to have its own vocabularies which mainly impact the quality of the HIR process results. Keywords queries are frequently defined using several lexical variants thus leading to the Vocabulary Mismatch phenomena (VM). Hence, more rigorous efforts are required in the health information retrieval domain in general and in query formulation especially to assist either clinicians’ practitioners or non-specialist during their daily retrieval practice. In this paper, we propose a novel approach that combines the clinical diagnosis generation via deep image captioning, Unified Medical Language System unique concept detection to get the best query formulation combination for an enhanced query. To ensure an efficient HIR process we refine the query by the user. Experimental results procured for the well-used search engines such as Google and Bing showed that when we used our D-MiQ query formulation, results effectiveness is considerably improved. Funding demonstrates that the utilization of merging various techniques is a valuable addition to the Query Expansion (QE) process leading to significantly improved precision rate by over 68% and reduced VM in Health Information Retrieval.