Improving Pronoun Translation for Statistical Machine Translation

Machine Translation is a well–established field, yet the majority of current systems translate sentences in isolation, losing valuable contextual information from previously translated sentences in the discourse. One important type of contextual information concerns who or what a coreferring pronoun corefers to (i.e., its antecedent). Languages differ significantly in how they achieve coreference, and awareness of antecedents is important in choosing the correct pronoun. Disregarding a pronoun’s antecedent in translation can lead to inappropriate coreferring forms in the target text, seriously degrading a reader’s ability to understand it. This work assesses the extent to which source-language annotation of coreferring pronouns can improve English–Czech Statistical Machine Translation (SMT). As with previous attempts that use this method, the results show little improvement. This paper attempts to explain why and to provide insight into the factors affecting performance.