We have intentionally chosen not to employ stemming or
stop word removal at this stage of the experiments. There
are two primary reasons for this: 1) chat posts are sparse and
often an entire post may consist of what might be considered
non-content bearing words under other contexts, so we
wish to preserve this in the hope that even the non-content
bearing words, or specific morphologies of words might
tend to assist in grouping like content; and 2) follow-on
techniques such as WordNet hypernym augmentation provides
automatic stemming of words, so it is not necessary
to do so when building our initial connectivity matrix.