Classifying Author Intention for Writer Feedback in Related Work

Proceedings of Recent Advances in Natural Language Processing, pages 178–187, Varna, Bulgaria, Sep 2–4, 2019. https://doi.org/10.26615/978-954-452-056-4_021 178 Classifying Author Intention for Writer Feedback in Related Work Arlene Casey School of Informatics University of Edinburgh Edinburgh, UK a.j.casey@sms.ed.ac.uk Bonnie Webber School of Informatics University of Edinburgh Edinburgh, UK bonnie@inf.ed.ac.uk Dorota Głowacka Dept. of Computer Science University of Helsinki Helsinki, Finland glowacka@cs.helsinki.fi Abstract The ability to produce high-quality pub- lishable material is critical to academic success but many Post-Graduate students struggle to learn to do so. While recent years have seen an increase in tools de- signed to provide feedback on aspects of writing, one aspect that has so far been neglected is the Related Work section of academic research papers. To address this, we have trained a supervised classifier on a corpus of 94 Related Work sections and evaluated it against a manually annotated gold standard. The classifier uses novel features pertaining to citation types and co-reference, along with patterns found from studying Related Works. We show that these novel features contribute to clas- sifier performance with performance be- ing favourable compared to other similar works that classify author intentions and consider feedback for academic writing. 1 Introduction Argument structures are key in allowing an au- thor to construct a persuasive message that real- izes the author’s intention. The automatic identi- fication of such intentions has been shown to be a valuable resource in areas such as summarising information (Teufel and Moens, 2002; Cohan and Goharian, 2015), and understanding citation func- tion and sentiment (Teufel et al., 2006; Jurgens et al., 2018). Recent years have seen more aca- demic writing tools focused on content that use an understanding of expected author intentions to assist in feedback. This is an important resource for Post-Graduate (PG) students who struggle to gain the necessary skills in academic writing that are critical to their success (Aitchison et al., 2012; Paltridge and Starfield, 2007). Automating under- standing of author intentions is challenging as re- search articles, whilst classed in their own genre, are known to differ in content and linguistic style across disciplines (Hyland, 2008). Despite these challenges, previous work using author intentions has been successful in automat- ing writing feedback though largely focused on the Abstract (Feltrim et al., 2006) or the Introduction (Cotos and Pendar, 2016; Abel, 2018). One reason for this focus on a single section of a research pa- per is that each section has its own purpose, which encourages different linguistic practices. Existing tools may concentrate on the Introduction due to formative work done by Swales (1990). Swales was one of the first to recognise these intentions in academic writing calling them rhetoric intentions. Swales showed how linguistic patterns in the In- troduction could be matched to intentions, such as establishing a territory or defining the problem. One section of academic papers that has not yet been explored for automated content feed- back is the Related Work section. One particu- lar challenge students have when learning to write is to find …

Content truncated to 500 words