Back

A Light-weight Text Summarizer for Fast Access to Medical Evidence

Sarker, A.; Yang, Y.-C.; Al-Garadi, M. A.

2020-05-26 health informatics
10.1101/2020.05.22.20110742 medRxiv
Show abstract

The performances of current medical text summarization systems rely on resource-heavy domain-specific knowledge sources, and preprocessing methods (e.g., classification or deep learning) for deriving semantic information. Consequently, these systems are often difficult to customize, extend or deploy in low-resource settings, and are operationally slow. We propose a fast summarization system that can aid practitioners at point-of-care, and, thus, improve evidence-based healthcare. At runtime, our system utilizes similarity measurements derived from pre-trained domain-specific word embeddings in addition to simple features, rather than clunky knowledge bases and resource-heavy preprocessing. Automatic evaluation on a public dataset for evidence-based medicine shows that our systems performance, despite the simple implementation, is statistically comparable with the state-of-the-art.

Matching journals

The top 3 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.