Amplifying scientific paper´s abstract by leveraging data-weighted reconstruction

This paper considers the problem of amplifying scientific paper´s abstract by using its citation sentences. While scientific paper´s abstract is concise and subjective, its citation sentences, which can be regarded as one kind of comment, are redundant and objective. A summary combining the merits of those two resource is helpful to researchers. A data-weighted reconstruction approach is proposed to generate this summary and sentence´s weight is learned through a ranking algorithm over hypergraph of bibliographic network. The experimental results show that the proposed approach can achieve significantly better than several document summarization´s techniques.