Combining Symbolic and Connectionist Techniques for Coordination in Natural Language

This paper describes a hybrid symbolic/connectionist system which combines symbolic and connectionist techniques for the structural interpretation of noun phrases. Using coordination (grammatical structures with conjunctions like "and") as a representative problem for a whole class of attachment problems, we describe a system which integrates syntactic and semantic knowledge for parsing "real world" text from a scientific technical corpus. Our hybrid model consists of a symbolic chart parser for parsing noun phrases, a symbolic preference module for semantic expectations, and a connectionist backpropagation network for semantic coordination relationships. We show that a symbolic syntactic parser and a connectionist semantic memory model can interact for resolving coordination problems.

@InProceedings{Wer90, 
 	 author =  {Wermter, Stefan},  
 	 title = {Combining Symbolic and Connectionist Techniques for Coordination in Natural Language}, 
 	 booktitle = {Proceedings of the 14th German Workshop on Artificial Intelligence},
 	 number = {},
 	 volume = {},
 	 pages = {186--195},
 	 year = {1990},
 	 month = {},
 	 publisher = {},
 	 doi = {}, 
 }