Keyphrase Generation: A Text Summarization Struggle

Proceedings of NAACL-HLT 2019, pages 666–672 Minneapolis, Minnesota, June 2 - June 7, 2019. c⃝2019 Association for Computational Linguistics 666 Keyphrase Generation: A Text Summarization Struggle Erion C¸ ano Institute of Formal and Applied Linguistics, Charles University, Prague, Czech Republic cano@ufal.mff.cuni.cz Ondˇrej Bojar Institute of Formal and Applied Linguistics, Charles University, Prague, Czech Republic bojar@ufal.mff.cuni.cz Abstract Authors’ keyphrases assigned to scientific ar- ticles are essential for recognizing content and topic aspects. Most of the proposed supervised and unsupervised methods for keyphrase gen- eration are unable to produce terms that are valuable but do not appear in the text. In this paper, we explore the possibility of consider- ing the keyphrase string as an abstractive sum- mary of the title and the abstract. First, we collect, process and release a large dataset of scientific paper metadata that contains 2.2 mil- lion records. Then we experiment with pop- ular text summarization neural architectures. Despite using advanced deep learning models, large quantities of data and many days of com- putation, our systematic evaluation on four test datasets reveals that the explored text sum- marization methods could not produce bet- ter keyphrases than the simpler unsupervised methods, or the existing supervised ones. 1 Introduction A valuable concept for searching and categoriz- ing scientific papers in digital libraries is the keyphrase (we use keyphrase and keyword inter- changeably), a short set of one or few words that represent concepts. Scientific articles are com- monly annotated with keyphrases based on tax- onomies of concepts and the authors’ judgment. Finding keyphrases that best describe the contents of a document is thus essential and rewarding. Most of the proposed keyphrase extraction so- lutions tend to be unsupervised (Florescu and Caragea, 2017; Nguyen and Luong, 2010; Rose et al., 2010; Bougouin et al., 2013; Campos et al., 2018) and generate terms by selecting the most ap- propriate candidates, ranking the candidates based on several features and finally returning the top N. Another way is to utilize datasets of texts and keywords for training supervised models with lin- guistic or other features to predict if candidates are keywords or not (Witten et al., 1999; Turney, 2000; Medelyan, 2009; Hulth, 2003). All above methods propose N keyphrases for each article which are joined together with “,” (or other separator like “;”) to form the keyphrase string of that article. They suffer from various problems or discrepancies. First, they are unable to find an optimal value for N and require it as a preset parameter. Furthermore, semantic and syntactic properties of article phrases are analyzed separately. The meaning of paragraphs, sections or entire document is thus missed. Lastly, only phrases that do appear in the article are returned. Meng et al. (2017) recently proposed a deep su- pervised keyphrase generation solution trained on a big dataset. It successfully solves the last two problems above, but not the first one. Motivated by recent advances in neural ma- chine translation and abstractive text summariza- tion (Vaswani et al., 2017; Foster et al., 2018; Rush …

Content truncated to 500 words