Conventional n-best reranking techniques often suffer from the limited scope of the n-best list, which rules out many potentially good alternatives. We instead propose forest reranking, a method that reranks the packed forest of exponentially many parses. Although exact inference is intractable with non-local features, we present an approximate algorithm that makes discriminative training practical over the whole Treebank. Our final result, an F-score of 91.7, outperforms both 50-best and 100-best reranking baselines, and is better than any previously reported systems trained on the Treebank. This method can be viewed as an integration of discriminative reranking with traditional chart parsing based on dynamic programming.
I will also discuss some on-going work on applying forest reranking to machine translation.
--------------------
References:
Liang Huang. Forest Reranking: Discriminative Parsing with Non-Local Features. To Appear in Proceedings of ACL 2008. http://www.cis.upenn.edu/~lhuang3/forest-rerank.pdf