Unlike classical lexical overlap metrics such as BLEU, most current evaluation metrics for
machine translation (for example, COMET or BERTScore) are based on black-box large …

This chapter delves into the recent accountability tools tailored for the evolving landscape of
machine learning models for complex reasoning tasks. With the increasing integration of …

Unlike classical lexical overlap metrics such as BLEU, most current evaluation metrics (such
as BERTScore or MoverScore) are based on black-box language models such as BERT or …

We demonstrate the capabilities of two model-agnostic local post-hoc model interpretability
methods, namely breakDown (BD) and shapley (SHAP), to explain the predictions of a black …

In the language domain, as in other domains, neural explainability takes an ever more
important role, with feature attribution methods on the forefront. Many such methods require …

Although explainable artificial intelligence (XAI) has achieved remarkable developments in
recent years, there are few efforts have been devoted to the following problems, namely, i) …

In comparison to the interpretation of classification models, the explanation of sequence
generation models is also an important problem, however it has seen little attention. In this …

Generating free-text rationales is a promising step towards explainable NLP, yet evaluating
such rationales remains a challenge. Existing metrics have mostly focused on measuring the …

State-of-the-art Entity Matching approaches rely on transformer architectures, such as BERT,
for generating highly contextualized embeddings of terms. The embeddings are then used to …

Reasoning over spans of tokens from different parts of the input is essential for natural
language understanding (NLU) tasks such as fact-checking (FC), machine reading …