ClaNNet (Classifier Neural Network) is a self-organizing neural network, capable of learning the structure of sentences and classifying unknown works. ClaNNet induces sentence structures by learning a word adjacency mapping (in terms of categories). The network predicts the categories (or parts of speech) of unknown words from the cues provided by the sentences structure and the word context. In doing so, ClaNNet Learns some grammatical rules. ClaNNet can also backtrack several words of input to find a misclassified word. A word category may have multiple successors. In such cases of ambiguity, ClaNNet chooses one of the alternatives. It corrects itself if its choice is not supported by further data. Word representations in ClaNNet's lexicon are updated during learning.
Husain, F. and Juell, P. (1996). A Self-Organizing Neural Network for the classification of Unknown Words. In Anjaneyulu, K.S.R., Sasikumar, M. and Ramani, S. (Ed.s), 'Knowledge Based Computer Systems: Research and Applications', Narosa Publishing House, New Delhi, India.
S --> NP, VP NP --> DET N VP --> V NP DET --> the DET --> a N --> dog N --> man V --> bit The dog bit the man. NP --> DET UNKNOWN-N The cat bit the man.