Head-driven Phrase Structure Grammar (HPSG) has been found to benefit from joint training and decoding under a uniform formalism . However, decoding this unified grammar has a higher time complexity ($O(n^5)$) than decoding either form individually . We propose an improved head scorer that helpsachieve a novel performance-preserved parscher in $O$($n^3$) time complexity . We also present a more effective, morein-depth, and general work on HPSG parsing. We also explore the generalmethod of training an H PSG-based parscher from only a constituent or dependencyannotations in a multilingual scenario .
Author(s) : Zuchao Li, Junru Zhou, Hai Zhao, Kevin ParnowLinks : PDF - Abstract
Code :
Keywords : complexity - head - time - decoding - driven -
- Data Scientist with Python
- Mathematics for Machine Learning by Marc Peter Deisenroth, A. Aldo Faisal and Cheng Soon Ong
- Python for Everybody on Coursera
- Khan Academy Matrices
- Google Cloud Professional Data Engineer Specialization
- AI for Everyone by Andrew Ng
- Khan Academy Linear Algebra
- Kaggle Learning