Abstract, Rationale, Stance: A Joint Model for Scientific Claim Verification

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 3580–3586 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 3580 Abstract, Rationale, Stance: A Joint Model for Scientific Claim Verification Zhiwei Zhang1,2, Jiyi Li2∗, Fumiyo Fukumoto2 and Yanming Ye1 Hangzhou Dianzi University, Hangzhou, China1 University of Yamanashi, Kofu, Japan2 hduxiaozhi97@gmail.com, {jyli,fukumoto}@yamanashi.ac.jp yeym@hdu.edu.cn Abstract Scientific claim verification can help the re- searchers to easily find the target scientific pa- pers with the sentence evidence from a large corpus for the given claim. Some existing works propose pipeline models on the three tasks of abstract retrieval, rationale selection and stance prediction. Such works have the problems of error propagation among the mod- ules in the pipeline and lack of sharing valu- able information among modules. We thus pro- pose an approach, named as ARSJOINT, that jointly learns the modules for the three tasks with a machine reading comprehension frame- work by including claim information. In ad- dition, we enhance the information exchanges and constraints among tasks by proposing a regularization term between the sentence atten- tion scores of abstract retrieval and the esti- mated outputs of rational selection. The exper- imental results on the benchmark dataset SCI- FACT show that our approach outperforms the existing works. 1 Introduction A system of scientific claim verification can help the researchers to easily find the target scientific papers with the sentence evidence from a large corpus for the given claim. To address this issue, Wadden et al. (2020) introduced scientific claim verification which consists of three tasks. As il- lustrated in Figure 1, for a given claim, the sys- tem finds the abstracts which are related to the claim from a scholarly document corpus (abstract retrieval task); it selects the sentences which are the evidences in the abstract related to the claim (rationale selection task); it also classifies whether the abstract/sentences support or refute the claims (stance prediction task). Wadden et al. (2020) also provided a dataset called SCIFACT. Most of the existing works of general claim ver- ification are based on pipeline models (Soleimani et al., 2020; Alonso-Reina et al., 2019; Liu et al., Figure 1: An example of scientific claim verification. 2020; Zhou et al., 2019; Nie et al., 2019; Lee et al., 2020b); some works utilize joint optimization strategies (Lu and Li, 2020; Yin and Roth, 2018; Hidey et al., 2020). These models attempted to jointly optimize the rationale selection and stance prediction, but did not directly link the two mod- ules (Li et al., 2020). In the case of the scientific claim verification, Wadden et al. (2020) proposed a baseline model VERISCI based on a pipeline of three components for the three tasks. Pradeep et al. (2021) proposed a pipeline model called VERT5ERINI which utilized the pre-trained lan- guage model T5 (Raffel et al., 2020) and adapted a pre-trained sequence-to-sequence model. Li et al. (2020) jointly trained two tasks of rationale selec- tion and stance prediction, and had a pipeline on abstract retrieval task and the joint module. …

Content truncated to 500 words