Uncertainty estimation is important for ensuring safety and robustness of AI systems. While most research in the area has focused on un-structured prediction tasks, limited work has investigated general uncertainty estimation approaches for structured prediction. Thus, this work aims to investigate uncertainty estimation for structured prediction tasks within a single unified and interpretable probabilistic ensemble-based framework. We consider: uncertainty estimation for sequence data at the token-level and complete sequence-level; interpretations for, and applications of, various measures of uncertainty; and discuss both the theoretical and practical challenges associated with obtaining them. This work also provides baselines for token-level and sequence-level error detection, and sequence-level out-of-domain input detection on the WMT’14 English-French and WMT’17 English-German translation and LibriSpeech speech recognition datasets.

Andrey Malinin

Mark Gales

Language is one of the key forms of communication. We study methods of language representation and understanding to simplify human-computer interactions.

Natural language processing 

Distributional shift is the mismatch between training and deployment data that is ubiquitous in the real-world. Studying this phenomenon can enable safer and more reliable ML systems.

Distributional shift

Uncertainty estimation enables detecting when ML models make mistakes. This is of critical importance in high risk machine learning applications, such as autonomous vehicle and medical ML.

Uncertainty estimation 

Language barriers hinder global communication and access to worldwide knowledge. By improving machine translation systems, we hope to facilitate the exchange of culture and information.

Machine translation

Speech is an important data modality and relatives to applications such as speech recognition and speech synthesis, which are core technologies in products such as vocal assistants.

Uncertainty Estimation in Autoregressive Structured Prediction