Counterfactual Learning from Bandit Feedback under Deterministic Logging: A Case Study in Statistical Machine Translation
This addresses a practical challenge for commercial SMT systems by enabling optimization from limited, deterministic logs, though it is incremental as it builds on existing counterfactual learning methods.
The paper tackles the problem of counterfactual learning for statistical machine translation from deterministic bandit logs, where commercial systems log only the most probable translations, and shows that smoothing out deterministic components with control variates enables learning, resulting in improvements of up to 2 BLEU points in simulations.
The goal of counterfactual learning for statistical machine translation (SMT) is to optimize a target SMT system from logged data that consist of user feedback to translations that were predicted by another, historic SMT system. A challenge arises by the fact that risk-averse commercial SMT systems deterministically log the most probable translation. The lack of sufficient exploration of the SMT output space seemingly contradicts the theoretical requirements for counterfactual learning. We show that counterfactual learning from deterministic bandit logs is possible nevertheless by smoothing out deterministic components in learning. This can be achieved by additive and multiplicative control variates that avoid degenerate behavior in empirical risk minimization. Our simulation experiments show improvements of up to 2 BLEU points by counterfactual learning from deterministic bandit feedback.