Pronoun Translation in English-French Machine Translation: An Analysis of Error Types

TL;DR

This paper analyzes the challenges of pronoun translation in English-French machine translation, comparing rule-based, statistical, and neural systems, revealing strengths and weaknesses with detailed error analysis.

Contribution

It provides a comprehensive evaluation of different MT systems on pronoun translation using the PROTEST test suite, highlighting specific error types and areas for improvement.

Findings

Rule-based systems perform poorly due to oversimplification

SMT and early NMT systems lack awareness of pronoun properties

Transformer-based NMT shows promising results but struggles with cross-sentence dependencies

Abstract

Pronouns are a long-standing challenge in machine translation. We present a study of the performance of a range of rule-based, statistical and neural MT systems on pronoun translation based on an extensive manual evaluation using the PROTEST test suite, which enables a fine-grained analysis of different pronoun types and sheds light on the difficulties of the task. We find that the rule-based approaches in our corpus perform poorly as a result of oversimplification, whereas SMT and early NMT systems exhibit significant shortcomings due to a lack of awareness of the functional and referential properties of pronouns. A recent Transformer-based NMT system with cross-sentence context shows very promising results on non-anaphoric pronouns and intra-sentential anaphora, but there is still considerable room for improvement in examples with cross-sentence dependencies.