Enhance Topics Analysis based on Keywords Properties
This work addresses the problem of topic model evaluation for researchers and practitioners, offering an incremental improvement over existing methods.
The paper tackles the challenge of evaluating topic models without gold-standard topics by introducing a specificity score based on keyword properties to select the most informative topics, showing in experiments that it compresses state-of-the-art results with much lower information loss than coherence-based methods.
Topic Modelling is one of the most prevalent text analysis technique used to explore and retrieve collection of documents. The evaluation of the topic model algorithms is still a very challenging tasks due to the absence of gold-standard list of topics to compare against for every corpus. In this work, we present a specificity score based on keywords properties that is able to select the most informative topics. This approach helps the user to focus on the most informative topics. In the experiments, we show that we are able to compress the state-of-the-art topic modelling results of different factors with an information loss that is much lower than the solution based on the recent coherence score presented in literature.