Learning internal representations by error propagation
Parallel distributed processing: explorations in the microstructure of cognition, vol. 1
Connectionist, Statistical, and Symbolic Approaches to Learning for Natural Language Processing
Connectionist, Statistical, and Symbolic Approaches to Learning for Natural Language Processing
Using hybrid connectionist learning for speech/language analysis
Connectionist, Statistical, and Symbolic Approaches to Learning for Natural Language Processing
Integrating different learning approaches into a multilingual spoken language translation system
Connectionist, Statistical, and Symbolic Approaches to Learning for Natural Language Processing
Learning dialog act processing
COLING '96 Proceedings of the 16th conference on Computational linguistics - Volume 2
Learning dialog act processing
COLING '96 Proceedings of the 16th conference on Computational linguistics - Volume 2
Journal of Artificial Intelligence Research
| Hi-index | 4.10 |
We have developed a hybrid connectionist/symbolic spoken-language processing system.The SCREEN (Symbolic Connectionist Robust Enterprise for Natural Language) system has to be sufficiently robust to analyze real-world spoken language, which is unpredictable and full of errors. SCREEN uses as its input word hypotheses generated by a speech recognizer, which identifies spoken words for spoken-language processing. After receiving word hypotheses, the system produces many utterance hypotheses and then determines which are most likely to be accurate. The system is robust enough to continue processing spoken language even when encountering incorrect word hypotheses produced by speech recognizers. SCREEN learns a flat syntax, semantics, and pragmatics representation, and deals with uncommon syntactic and semantic language irregularities. It can be used for relatively large and complex tasks. Early connectionist networks could be used only for relatively small tasks.