Automated fact-checking systems verify claims against evidence to predict their veracity. In real-world scenarios, the retrieved evidence may not unambiguously support or refute the claim and yield conflicting but valid interpretations. Existing fact-checking datasets assume that the models developed with them predict a single veracity label for each claim, thus discouraging the handling of such ambiguity. To address this issue we present AmbiFC,1 a fact-checking dataset with 10k claims derived from real-world information needs. It contains fine-grained evidence annotations of 50k passages from 5k Wikipedia pages. We analyze the disagreements arising from ambiguity when comparing claims against evidence in AmbiFC, observing a strong correlation of annotator disagreement with linguistic phenomena such as underspecification and probabilistic reasoning. We develop models for predicting veracity handling this ambiguity via soft labels, and find that a pipeline that learns the label distribution for sentence-level evidence selection and veracity prediction yields the best performance. We compare models trained on different subsets of AmbiFC and show that models trained on the ambiguous instances perform better when faced with the identified linguistic phenomena.

This paper investigates explainability in Natural Legal Language Processing (NLLP). We study the task of legal outcome prediction of the European Court of Human Rights cases in a ternary classification setup, where a language model is fine-tuned to predict whether an article has been claimed and violated (positive outcome), claimed but not violated (negative outcome) or not claimed at all (null outcome). Specifically, we experiment with three popular NLP explainability methods. Correlating the attribution scores of input-level methods (Integrated Gradients and Contrastive Explanations) with rationales from court rulings, we show that the correlations are very weak, with absolute values of Spearman and Kendall correlation coefficients ranging between 0.003 and 0.094. Furthermore, we use a concept-level interpretability method (Concept Erasure) with human expert annotations of legal reasoning, to show that obscuring legal concepts from the model representation has an insignificant effect on model performance (at most a decline of 0.26 F1). Therefore, our results indicate that automated legal outcome prediction models are not reliably grounded in legal reasoning.

Many NLP tasks have benefited from transferring knowledge from contextualized word embeddings, however the picture of what type of knowledge is transferred is incomplete. This paper studies the types of linguistic phenomena accounted for by language models in the context of a Conversational Question Answering (CoQA) task. We identify the problematic areas for the finetuned RoBERTa, BERT and DistilBERT models through systematic error analysis - basic arithmetic (counting phrases), compositional semantics (negation and Semantic Role Labeling), and lexical semantics (surprisal and antonymy). When enhanced with the relevant linguistic knowledge through multitask learning, the models improve in performance. Ensembles of the enhanced models yield a boost between 2.2 and 2.7 points in F1 score overall, and up to 42.1 points in F1 on the hardest question classes. The results show differences in ability to represent compositional and lexical information between RoBERTa, BERT and DistilBERT.

When a reader is first introduced to an entity, its referring expression must describe the entity. For entities that are widely known, a single word or phrase often suffices. This paper presents the first study of how expressions that refer to the same entity develop over time. We track thousands of person and organization entities over 20 years of New York Times (NYT). As entities move from hearer-new (first introduction to the NYT audience) to hearer-old (common knowledge) status, we show empirically that the referring expressions along this trajectory depend on the type of the entity, and exhibit linguistic properties related to becoming common knowledge (e.g., shorter length, less use of appositives, more definiteness). These properties can also be used to build a model to predict how long it will take for an entity to reach hearer-old status. Our results reach 10-30% absolute improvement over a majority-class baseline.

The current paper covers several strategies we used to ‘break’ predictions of sentiment analysis systems participating in the BLGNLP2017 workshop. Specifically, we identify difficulties of participating systems in understanding modals, subjective judgments, world-knowledge based references and certain differences in syntax and perspective.