either have a set of pre-defined topic labels (e.g provided by Historians) and we could try to run ZeSTE / ProZe experiments in order to get a fine grained classification of text documents according to these topics;
or run a pure topic modeling algorithm (via ToModAPI) over a collection and see what topics are being formed, the difficulty being to estimate the right number of topics
It would be interesting to include somehow the topic (e.g.
schema:about
) of the smell, e.g. war, odour of sanctity...