Connectionist Context Processing for Phrase Filtering

This paper describes connectionist recurrent plausibility networks that are examined for learning a scanning understanding of a large number of “real-world” phrases. These plausibility networks encode incremental context of phrases in recurrent connections and therefore allow for filtering phrases according to different context classes. Since input representations are automatically acquired as significance vectors these plausibility networks support the portability across different domains. Furthermore since underlying regularities for context assignment are learned from an unrestricted corpus of phrases these networks support adaptability and robust processing independent of the underlying grammatical constructions.

@InProceedings{Wer93a,
 	 author =  {Wermter, Stefan},
 	 title = {Connectionist Context Processing for Phrase Filtering},
 	 booktitle = {Proceedings of the World Congress on Neural Networks},
 	 journal = {None},
 	 editors = {}
 	 number = {}
 	 volume = {}
 	 pages = {100--103},
 	 year = {1993},
 	 month = {Jan},
 	 publisher = {Lawrence Erlbaum Associates},
 	 doi = {}
 }