Comparison of Automatic Extraction of Research Highlights and Abstracts of Journal Articles
Wai Ming Wang, Eric W.K. See-To, Hong Tao Lin, Zhi Li · Proceedings of the 2nd International Conference on Computer Science and Application Engineering · 2018
Many academic journals1 require authors to submit a list of bullet points, called highlights. Highlights provide concise information for readers to review the article quickly. However, some journals and old articles do not have the highlights. It is useful to develop an automated method for extracting highlights of these articles. In this paper, we study the important features of research highlight extraction and identify the differences between extraction of highlights and abstracts of journal articles. We use information science related journal articles as test data. We quantitatively evaluate 23 common unsupervised extractive text summarization methods. The results show that the application of extractive text summarization is suitable for research highlight extraction. In particular, TextRank obtains the highest recall and the title word method achieves the highest precision. The result could help researchers developing new methods of highlight extraction.