Supervised Automatic Sentence Parsing For Hadiyaa Language

Abstract

The Syntactic Processing (parsing) is the method for understanding how words and phrases are combined to synthesize a sentence whether it fits to the given language grammar structure. Parsing is an essential intermediate phase for the building language semantic analysis systems. This research is mainly concerned with developing prototype to automatic sentence parsing for Hadiyya Language using hybrid approach. Thus, to conduct this research Rule-Based Learning Approach with Chart parsing algorithm combined with automatic feature construction algorithms (Maximum Entropy model) used to realize the parser. In this thesis a modified version of the Chart parsing algorithm was applied to a corpus that was manually annotated with a skeletal syntactic bracketing which confirms to the Penn Treebank POS Tags. The Hadiyya language parsing system prototype is evaluated with PARSEVAL metrics, which computed by Precision, Recall and F-measure. The experment conducted on the total corpus of 549 sentences colections. The obtained results were 73.0 % precision and Recall value of 64.3 % on test data set. The higher precision value when we compared with recall showed us the parser is effective in syntactic analysis.