Analyzing Statistical and Syntactical English Text for Word Prediction and Text Generation

摘要

This research describes a technique for word and phrase prediction to help the writer of an English text in text generation and to speed up the typing process. The technique consists of a learning phase and a generation or prediction phase. The learning phase learns the English phrase and sentence syntax as well as keeps all the corpora for reference against the syntax and the actual text. The prediction of the next word or phrase is based on the preceding one or more words and the history and frequency of occurrences of phrase mappings. This research establishes that word and phrase prediction based on the preceding word or words depending on statistical or frequency use of the words is enhanced by coupling it with syntax structures to ensure that a form of semantics and grammar correctness is maintained. The approach shows the way to learn both the words strings and the syntax structures which are later used for the prediction or generation purposes. We present results to demonstrate that our approach gives enhancements to key stroke reduction and a more accurate and grammatically correct prediction.