Topic Labelling

Topic labelling, in general, is a framework aiming to automatically extract topics from given text documents. Each extracted topic is typically represented by the top n words which have the highest \(P(w_i|t_j)\), i.e. Generative probabilities of word \(w_i\), given the topic \(t_j\).

Topic Evolution

Topic evolution (Dynamic Topic Models), is different from topic labelling by leveraging the knowledge of different documents belonging to a different time-slice in an attempt to map how the words in a topic change over time.

At each time-slice, the output for topic evolution would be similar as topic labelling: topics and their corresponding topic words. The model also have the topic for all time-slices by finding semantically similar documents. So we could see how topic words changes during the time.

TOPIC MODELLING VISUALISATION

Topic Labelling

Topic labelling, in general, is a framework aiming to automatically extract topics from given text documents. Each extracted topic is typically represented by the top n words which have the highest \(P(w_i|t_j)\), i.e. Generative probabilities of word \(w_i\), given the topic \(t_j\).

Topic Evolution

Topic evolution (Dynamic Topic Models), is different from topic labelling by leveraging the knowledge of different documents belonging to a different time-slice in an attempt to map how the words in a topic change over time.

© 2020 Copyright: NLP Group, University of Warwick