Query-focused summarisation in research articles based on semantic function of sentences

Query-focused summarisation (QFS) in research articles is usually used to help researchers to sum up content related to specific aspects of research articles. However, most QFS approaches failed to consider the inherent structure of research articles to speculate the semantic functions of sentences to make summarisations relate to the given aspect more precisely. Systematic functional linguistic studies suggested that research articles contain inherent structures in which sections and sentences have their specific functions. We suppose these structures can be used as auxiliary information for scientific summarisation. In this article, we seek to improve existing extractive QFS methods by using the macrostructure and discourse segment structure of research articles. We categorise sentences in research articles into different types according to their semantic functions and assign different weights to each type of sentence using an action-based relevance calculation method. We show that our system outperforms the baseline system on a benchmark dataset. Our findings suggest that using the inherent structure of research articles as assistance is practical for scientific summarisation.