Learning the Work of a Librarian: a Connectionist Model for Semantic Classification

In this paper we describe a recurrent sequential model for classifying book titles based on an existing library classification. Sequential classification networks learn and generalize semantically and syntactically unrestricted book titles from several library classes. The feature representation of words is acquired automatically since the feature representation is grounded in corpus occurrence. The performance of the model demonstrates that sequential classification networks grounded in corpus occurrence can scale up semantic classification at a "scanning" level of natural language understanding.

@InProceedings{Wer91a,
 	 author =  {Wermter, Stefan},
 	 title = {Learning the Work of a Librarian: a Connectionist Model for Semantic Classification},
 	 booktitle = {Proceedings of the AAAI Spring Symposium on Connectionist Natural Language Processing},
 	 journal = {None},
 	 editors = {}
 	 number = {}
 	 volume = {}
 	 pages = {72--77},
 	 year = {1991},
 	 month = {}
 	 publisher = {None},
 	 doi = {}
 	 url = {None},
 }