A Hybrid Approach to Parsing Natural Languages

Ambiguities in natural languages make processing (parsing) them a difficult task. Parsing is even more difficult when dealing with a structurally complex natural language such as Arabic. In this paper, we briefly highlight some of the complex structure of Arabic, and we identify different parsing approaches and briefly discuss their limitations. Our goal is to produce a hybrid parser, by combining different parsing approaches, which retains the advantages of data-driven approaches but is guided by a set of grammatical rules to produce more accurate results. We describe a novel technique for directly combining different parsing approaches. Results for our initial experiments that we have conducted in this work, and our plans for future work are also presented.