Combinatorial Topology-based Semantic Clustering Applied to PubMed

To confront with an ever increasing number of published scientific articles, an effective, efficient, and easy-to-use tool is required to support biomedical scientists, while entering a new scientific field and encountering clinical decision, to organize a vast amount of PubMed abstracts into the panorama of specific topics according to their relevance. In brief, the set of associations among frequently co-occurring terms in given a set of PubMed documents forms naturally a simplicial complex. Afterwards each connected component of this simplicial complex represents a concept in the collection. Based on these concepts, documents can be clustered into meaningful classes. This paper presents an alternative search engine that applies a combinatorial topological method to automatically extract semantic clusters from the PubMed database of biomedical literature. We use several qualitative parameters to perform the user study that shows users are able to reduce search time. This clustering search engine is publicly available at http://ginni. bme.ntu. edu. tw/.