BACO: A Background Knowledge- and Content-Based Framework for Citing Sentence Generation

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 1466–1478 August 1–6, 2021. ©2021 Association for Computational Linguistics 1466 BACO: A Background Knowledge- and Content-Based Framework for Citing Sentence Generation Yubin Ge1, Ly Dinh1, Xiaofeng Liu2, Jinsong Su3, Ziyao Lu3, Ante Wang3, Jana Diesner1 1University of Illinois at Urbana-Champaign, USA 2Harvard University, USA 3Xiamen University, China {yubinge2, dinh4, jdiesner}@illinois.edu jssu@xmu.edu.cn Abstract In this paper, we focus on the problem of citing sentence generation, which entails gen- erating a short text to capture the salient in- formation in a cited paper and the connec- tion between the citing and cited paper. We present BACO, a BAckground knowledge- and COntent-based framework for citing sen- tence generation, which considers two types of information: (1) background knowledge by leveraging structural information from a cita- tion network; and (2) content, which repre- sents in-depth information about what to cite and why to cite. First, a citation network is en- coded to provide background knowledge. Sec- ond, we apply salience estimation to identify what to cite by estimating the importance of sentences in the cited paper. During the decod- ing stage, both types of information are com- bined to facilitate the text generation. We then conduct joint training of the generator and cita- tion function classification to make the model aware of why to cite. Our experimental results show that our framework outperforms compar- ative baselines. 1 Introduction A citation systematically, strategically, and criti- cally synthesizes content from a cited paper in the context of a citing paper (Smith, 1981). A paper’s text that refers to prior work, which we herein refer to as citing sentences, forms the conceptual basis for a research question or problem; identifies is- sues, contradictions, or gaps with state of the art solutions; and prepares readers to understand the contributions of a citing paper, e.g., in terms of theory, methods, or findings (Elkiss et al., 2008). Writing meaningful and concise citing sentences that capture the gist of cited papers and identify connections between citing and cited papers is not trivial (White, 2004). Learning how to write up in- formation about related work with appropriate and meaningful citations is particularly challenging for new scholars (Mansourizadeh and Ahmad, 2011). To assist scholars with note taking on prior work when working on a new research problem, this paper focuses on the task of citing sentence genera- tion, which entails identifying salient information from cited papers and capturing connections be- tween cited and citing papers. With this work, we hope to reduce scientific information overload for researchers by providing examples of concise cit- ing sentences that address information from cited papers in the context of a new research problem and related write up. While this task cannot and is not meant to replace the scholarly tasks of finding, reading, and synthesizing prior work, the proposed computational solution is intended to support espe- cially new researchers in practicing the process of writing …

Content truncated to 500 words