Paper: Latent Anaphora Resolution for Cross-Lingual Pronoun Prediction

ACL ID D13-1037
Title Latent Anaphora Resolution for Cross-Lingual Pronoun Prediction
Venue Conference on Empirical Methods in Natural Language Processing
Session Main Conference
Year 2013

This paper addresses the task of predicting the correct French translations of third-person sub- ject pronouns in English discourse, a problem that is relevant as a prerequisite for machine translation and that requires anaphora resolu- tion. We present an approach based on neu- ral networks that models anaphoric links as latent variables and show that its performance is competitive with that of a system with sep- arate anaphora resolution while not requiring any coreference-annotated training data. This demonstrates that the information contained in parallel bitexts can successfully be used to ac- quire knowledge about pronominal anaphora in an unsupervised way. 1 Motivation When texts are translated from one language into another, the translation reconstructs the meaning or function of t...