We investigated the performance efficacy of beam search parsing and deep parsing techniques in probabilistic HPSG parsing using the Penn treebank. We first tested the beam thresholding and iterative parsing developed for PCFG parsing with an HPSG. Next, we tested three techniques originally developed for deep parsing: quick check, large constituent inhibition, and hybrid parsing with a CFG chunk parser. The contributions of the large constituent inhibition and global thresholding were not significant, while the quick check and chunk parser greatly contributed to total parsing performance. The precision, recall and average parsing time for the Penn treebank (Section 23) were 87.85%, 86.85%, and 360 ms, respectively. 1

We present a novel corpus-driven approach towards grammar approximation for a linguistically deep Head-driven Phrase Structure Grammar. With an unlexicalized probabilistic context-free grammar obtained by Maximum Likelihood Estimate on a largescale automatically annotated corpus, we are able to achieve parsing accuracy higher than the original HPSG-based model. Different ways of enriching the annotations carried by the approximating PCFG are proposed and compared. Comparison to the state-of-the-art latent-variable PCFG shows that our approach is more suitable for the grammar approximation task where training data can be acquired automatically. The best approximating PCFG achieved ParsEval F1 accuracy of 84.13%. The high robustness of the PCFG suggests it is a viable way of achieving full coverage parsing with the hand-written deep linguistic grammars. 1