MultiGBS: A multi-layer graph approach to biomedical summarization

Automatic text summarization methods generate a shorter version of the input to assist reader in gaining quick yet informative gist. Existing generally focus on single aspect when selecting sentences, causing potential loss essential information. In this study, we propose domain-specific method that models document as multi-layer graph enable multiple features be processed at same time. The used paper are word similarity, semantic and co-reference which modelled three different layers. unsupervised selects sentences from based MultiRank algorithm number concepts. proposed MultiGBS employs UMLS extracts concepts relationships using tools such SemRep, MetaMap, OGER. Extensive evaluation by ROUGE BERTScore shows increased F-measure values.