In this work, we study hallucinations in Neural Machine Translation (NMT), which lie at an extreme end on the spectrum of NMT pathologies. Firstly, we connect the phenomenon of hallucinations under source perturbation to the Long-Tail theory of Feldman, and present an empirically validated hypothesis that explains hallucinations under source perturbation. Secondly, we consider hallucinations under corpus-level noise (without any source perturbation) and demonstrate that two prominent types of natural hallucinations (detached and oscillatory outputs) could be generated and explained through specific corpus-level noise patterns. Finally, we elucidate the phenomenon of hallucination amplification in popular data-generation processes such as Backtranslation and sequence-level Knowledge Distillation. We have released the datasets and code to replicate our results.

Automatic metrics are commonly used as the exclusive tool for declaring the superiority of one machine translation system’s quality over another. The community choice of automatic metric guides research directions and industrial developments by deciding which models are deemed better. Evaluating metrics correlations with sets of human judgements has been limited by the size of these sets. In this paper, we corroborate how reliable metrics are in contrast to human judgements on – to the best of our knowledge – the largest collection of judgements reported in the literature. Arguably, pairwise rankings of two systems are the most common evaluation tasks in research or deployment scenarios. Taking human judgement as a gold standard, we investigate which metrics have the highest accuracy in predicting translation quality rankings for such system pairs. Furthermore, we evaluate the performance of various metrics across different language pairs and domains. Lastly, we show that the sole use of BLEU impeded the development of improved models leading to bad deployment decisions. We release the collection of 2.3M sentence-level human judgements for 4380 systems for further analysis and replication of our work.

We describe a novel approach to machine translation that combines the strengths of the two leading corpus-based approaches: Phrasal SMT and EBMT. We use a syntactically informed decoder and reordering model based on the source dependency tree, in combination with conventional SMT models to incorporate the power of phrasal SMT with the linguistic generality available in a parser. We show that this approach significantly outperforms a leading string-based Phrasal SMT decoder and an EBMT system. We present results from two radically different language pairs, and investigate the sensitivity of this approach to parse quality by using two distinct parsers and oracle experiments. We also validate our automated BLEU scores with a small human evaluation.

One of the problems facing translation systems that automatically extract transfer mappings (rules or examples) from bilingual corpora is the trade-off between contextual specificity and general applicability of the mappings, which typically results in conflicting mappings without distinguishing context. We present a machine-learning approach to choosing between such mappings, using classifiers that, in effect, selectively expand the context for these mappings using features available in a linguistic representation of the source language input. We show that using these classifiers in our machine translation system significantly improves the quality of the translated output. Additionally, the set of distinguishing features selected by the classifiers provides insight into the relative importance of the various linguistic features in choosing the correct contextual translation.

We describe MSR-MT, a large-scale example-based machine translation system under development for several language pairs. Trained on aligned English-Spanish technical prose, a blind evaluation shows that MSR-MT’s integration of rule-based parsers, example based processing, and statistical techniques produces translations whose quality in this domain exceeds that of uncustomized commercial MT systems.

Translation systems that automatically extract transfer mappings (rules or examples) from bilingual corpora have been hampered by the difficulty of achieving accurate alignment and acquiring high quality mappings. We describe an algorithm that uses a best-first strategy and a small alignment grammar to significantly improve the quality of the mappings extracted. For each mapping, frequencies are computed and sufficient context is retained to distinguish competing mappings during translation. Variants of the algorithm are run against a corpus containing 200K sentence pairs and evaluated based on the quality of resulting translations.