Paper: Designing Agreement Features for Realization Ranking

ACL ID C10-2119
Title Designing Agreement Features for Realization Ranking
Venue International Conference on Computational Linguistics
Session Poster Session
Year 2010

This paper shows that incorporating lin- guistically motivated features to ensure correct animacy and number agreement in an averaged perceptron ranking model for CCG realization helps improve a state-of- the-art baseline even further. Tradition- ally, these features have been modelled us- ing hard constraints in the grammar. How- ever, given the graded nature of grammat- icality judgements in the case of animacy we argue a case for the use of a statisti- cal model to rank competing preferences. Though subject-verb agreement is gener- ally viewed to be syntactic in nature, a pe- rusal of relevant examples discussed in the theoretical linguistics literature (Kathol, 1999; Pollard and Sag, 1994) points to- ward the heterogeneous nature of English agreement. Compared to writing gram- mar rule...