future_work.tex

\section{Future Work}
\label{sec:future_work}
Though many classifiers have been attempted, there is room to improve and combine them in future extensions to this work.

The ensemble classification approach could be expanded to include more classifiers and some additional
features so that it could learn which classifier to prefer depending
upon the features of the input sentence, effectively utilising "expert" classifiers.

Additionally, the current probabilistic model assumes that words
occur in a sentence independently. This is clearly not the case, and is just a useful simplification.
One way to extend the model would be to consider closer approximations to the
true distribution, for example by considering pairs of words instead of words on
their own or by considering a joint distribution with a sentiment or topic
analysis. This would also allow the probabilistic models to have an appreciation of word context, which is currently not present in that classifier.