A time-aware approach for boosting medical records search

Abstract

Medical records are collections of documents recording a patient's changing conditions, exhibiting temporal characteristic. Yet previous works on medical records search did not pay attention to it. We propose to model the medical records as sequential data, and utilize the temporal similarity between them to improve the performance of medical records search. In this paper, we propose a Temporal Bag-of-Words model to represent medical records as document sequence. In which framework, we adopt Dynamic Time Warping algorithm to calculate the temporal similarity between sequences. Then a clustering-based combination method is proposed for re-ranking. Experiments on TREC Medical Track data shows the effectiveness of the proposed framework for boosting medical records search.