{"type":"ScholarlyArticle","title":"On Guaranteed Optimal Robust Explanations for NLP Models","authors":[{"name":"Emanuele La Malfa","slug":"emanuele-la-malfa","given":"Emanuele","family":"La Malfa","affiliation":""},{"name":"Rhiannon Michelmore","slug":"rhiannon-michelmore","given":"Rhiannon","family":"Michelmore","affiliation":""},{"name":"Agnieszka M. Zbrzezny","slug":"agnieszka-m-zbrzezny","given":"Agnieszka M.","family":"Zbrzezny","affiliation":""},{"name":"Nicola Paoletti","slug":"nicola-paoletti","given":"Nicola","family":"Paoletti","affiliation":""},{"name":"Marta Kwiatkowska","slug":"marta-kwiatkowska","given":"Marta","family":"Kwiatkowska","affiliation":""}],"abstract":"We build on abduction-based explanations for machine learning and develop a method for computing local explanations for neural network models in natural language processing (NLP). Our explanations comprise a subset of the words of the input text that satisfies two key features: optimality w.r.t. a user-defined cost function, such as the length of explanation, and robustness, in that they ensure prediction invariance for any bounded perturbation in the embedding space of the left-out words. We present two solution algorithms, respectively based on implicit hitting sets and maximum universal subsets, introducing a number of algorithmic improvements to speed up convergence of hard instances. We show how our method can be configured with different perturbation sets in the embedded space and used to detect bias in predictions by enforcing include/exclude constraints on biased terms, as well as to enhance existing heuristic-based NLP explanation frameworks such as Anchors. We evaluate our framework on three widely used sentiment analysis tasks and texts of up to 100 words from SST, Twitter and IMDB datasets, demonstrating the effectiveness of the derived explanations.","doi":"10.24963/ijcai.2021/366","url":"https://ijcai.org/proceedings/2021/366","pdf":"https://ijcai.org/proceedings/2021/0366.pdf","markdown":"https://ijcai.org/proceedings/2021/366.md","bibtex":"https://ijcai.org/proceedings/2021/bibtex/366","volume":{"slug":"2021","title":"Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence","short_title":"IJCAI 2021","url":"https://ijcai.org/proceedings/2021/","year":2021,"location":"Montreal","dates":"19-27 August 2021","editors":"Prof. Zhi-Hua Zhou","isbn":"978-0-9992411-9-6","publisher":"International Joint Conferences on Artificial Intelligence Organization","doi_prefix":"10.24963/ijcai.2021"},"section":"Machine Learning","track":"Main Track","number":366,"pages":{"first":2658,"last":2665},"keywords":["Machine Learning: Adversarial Machine Learning","Machine Learning: Explainable/Interpretable Machine Learning","Natural Language Processing: Sentiment Analysis and Text Mining"],"videos":[],"references":[],"license_note":"Copyright International Joint Conferences on Artificial Intelligence."}