Word Prediction

Word Prediction

Speech and hand-writing recognition systems use prediction as well as measurement to interpret a signal. Previously recognised features are used to predict the current feature and this prediction is used to guide the analysis. The purpose of this project is to investigate efficient (i.e. fast and small) algorithms for predicting the next word and the next character in English text. The algorithm must be capable of being trained to support different vocabularies. One potential application might be in speeding up the entry of program text. The project will investigate fixed algorithms - the method is defined for a particular vocabulary and doesn't vary whatever text the user enters and adaptive algorithms - the method monitors the text enters and uses this to adjust its predictions. It may be useful to develop user-specific algorithms that can be trained to the vocabulary of a particular user.