Distinguishing different types of conference submissions: the ACL case study
Hong Wang, Barbara Di Eugenio, Shuyang Lin, Clement T. Yu · 2014
Many conferences in AI and NLP call for long and short papers; and satellite workshops co-locate with the main conference. In this work, we focus on distinguishing full from short from workshop papers, as submitted to some recent ACL conferences. We propose a framework that takes into account both metadata and content of the paper. To extract metadata, we devised a full-fledged paper parser. SVM models outperform the only previously published results by at least 3.6% as concerns distinguishing full from workshop papers. Metadata (number of tables/formulas), syntactic feature (syntactic complexity) and term TF-IDF score distinguish full from short papers, whereas the topic also distinguishes full from workshop papers.