Efficient HPSG Realization for Hybrid Machine Translation
Speaker:Stephan Oepen
Institution:University of Oslo and CSLI Stanford
Abstract:
I will review recent advances in grammar-based sentence realization from logical-form meaning representations. The LOGON MT prototype aims at the fully-automated, high-quality translation of Norwegian instructional texts (on backcountry activities) into English. The LOGON generator operates off underspecified meaning representations derived from `deep' grammatical analysis (in the LFG framework) and subsequent semantic transfer. The generator builds on the LinGO English Resource Grammar (in the HPSG framework) and combines a highly optimized chart-based algorithm with a rich, probabilistic model to rank alternate realizations. Integration of the stochastic model into the enumeration of outputs from the packed chart allows the generator to selectively unpack n-best lists of realizations with minimal search. Besides empirical results for the realization task when evaluated in isolation, I will present a summary of quantitative measures on the current development status (and promise) of the hybrid LOGON MT pipeline as a whole.