Improving verb phrase anaphora translation in English to French statistical machine translation

Abstract

This project investigates the translation of verb phrase anaphora within the source text of an English to French statistical machine translation system. VP anaphora is a common syntactic construction in English, although its translation into French is not always straightforward. I propose a method for improving VP anaphora translation by inserting a placeholder element in source sentences, and show that this method results in better translation of VP anaphora without affecting overall translation quality, both for clause-final VP ellipsis and for general VP anaphora constructions. I also implement a custom evaluation method for the analysis of VP ellipsis translation quality.