Toward Exploratory Search in Biomedicine: Evaluating Document Clusters by MeSH as a Semantic Anchor

TL;DR

This paper proposes a methodology for evaluating biomedical document clusters using MeSH terms as semantic anchors, aiming to enhance exploratory search capabilities for specialists.

Contribution

It introduces a novel evaluation framework incorporating human expertise and investigates the impact of full-text data on cluster quality.

Findings

MeSH-based semantic anchors improve cluster evaluation.

Full-text data enhances cluster quality.

Human expertise is crucial for effective evaluation.

Abstract

The current mode of biomedical literature search is severely limited in effectively finding information relevant to specialists. A potential approach to solving this problem is exploratory search, which allows users to interactively navigate through a vast document collection. As the first step toward exploratory search for specialists in biomedicine, this paper develops a methodology to evaluate quality of document clusters. For this purpose, we incorporate human expertise into data set creation and evaluation framework by leveraging MeSH terms as semantic anchors. In addition, we investigate the benefit of full-text data for improving cluster quality.