Advancing Improved Statistical and Syntactic Word Prediction
The internship’s broad objective is to improve the underlying mechanisms of Quillsoft’s flagship writing tool, WordQ, which is developed at Bloorview Kids Rehab, Canada’s largest children’s rehabilitation teaching hospital. WordQ software helps individuals to write by suggesting words and phrases based on what a user has already written. This is accomplished by a combination of statistics and grammar, which will be the focus of the intern’s activities. Currently, WordQ uses a short context window that precedes the current word to make its predictions. However, in a sentence fragment such as “The dog and cat went to the market and w‐“, the word “were” may be more likely than “was”, because the subject noun phrase is third person plural, but a short window would only consider the subsequence “market and w‐“, which is not sufficient. The intern’s chief activity will be the implementation of sentence‐wide grammatical constraints, as well as statistical methods for ranking alternative parses.