In many practical applications of topic modelling / thematic detection, there is a need to be able to handle incremental data, which involves being able to detect new topics, while still tracking the evolution of existing ones. At each new time window, one or more of the following could happen.
Tag new data to a existing topic
Form a new topic with new data
Split up an existing topic
Merge two existing topics
We want to develop a technique that allows stable modelling of topics of an evolving corpora over a series of (user-defined) time windows.
Description
In many practical applications of topic modelling / thematic detection, there is a need to be able to handle incremental data, which involves being able to detect new topics, while still tracking the evolution of existing ones. At each new time window, one or more of the following could happen.
We want to develop a technique that allows stable modelling of topics of an evolving corpora over a series of (user-defined) time windows.