Part I Analogy-based methods Skousen's analogical modelling algorithm: a comparison with lazy learning; Analogy, computation and linguistic theory; Constraints and preferences at work: an analogy-based approach to parsing grammatical relations. Part II Connectionist methods Towards a hybrid abstract generation system; Recurrent neural networks and natural language processing; Learning the semantics of aspect; Experiments in robust parsing with a guided propogation network; Learning semantic relationships and syntactic roles in a simple recurrent network. Part III Corpus-based methods A system for automating concordance line selection; A machine learning approach to anaphoric reference; Some methods for the extraction of bilingual terminology; Probabilistic part-of-speech tagging using decision trees; A new direction for sublanguage NLP; Efficient disambiguation by means of stochastic tree substitution grammars. Part IV Example-based machine translation Towards automatically aligning German compounds with English word groups; Corpus-based acquistion of transfer functions using psycholinguistic principles; A natural language translation neural network; Direct parse-tree translation in co-operation with the transfer method. Part V Statistical approaches Evaluating the information gain of probability based PP-disambiguation methods; Automatic error detection in part-of-speech tagging; A method of parsing English based on sentence form; A new approach to center tracking; Structuring raw discourse. Part VI Hybrid approaches Coarse-grained parallelism in natural language understanding: parsing as message passing; Parameterized message-passing for non-head-driven parsing; A stochastic government-binding parser; Evolutionary algorithms for dialogue optimization as an example of a hybrid nlp system; More for less: learning a wide covering grammar from a small training set. Part contents.
