Skip to content. | Skip to navigation

Personal tools
You are here: Home Research Advanced Community Information Systems (ACIS) D-VITA: Dynamic Visual Topic Analytics

D-VITA: Dynamic Visual Topic Analytics

D-VITA is a web based toolkit for interacting with visualizations of dynamic topic models, and for building & managing these models.

D-VITA is a web-based visual analytics toolkit for working with dynamic topic models [1]. Each topic model represents a statistical model of the evolution of topics in a text corpusses. Dynamic topic modeling was introduced by Blei and Lafferty [2], and there are several implementations available. The most recent evolution of D-VITA also offers experimental backend tools and user interfaces for convenient building of topic models using a toolchain metaphor [3].

D-VITA deploys a visual analytics metaphor for interaction with the topic model: topic evolution within the text corpus and word evolution within topics are visualized as theme rivers, topic distribution in documents is visualized as a pie chart, and for each topic and time slice the related documents in the corpus can be explored with a few mouse clicks. 

A document browser shows for each document in the corpus the most closely related other documents based on the topic model. The document browser also offers topic-based keyword search, i.e. finding documents based on the topics they expose.

Live Demos

There are three demonstration videos available right now. One short demo video of 5 minutes, and two longer, tutorial-style presentations of about 10 minutes each. See the embedded videos below:

Short Demo


Tutorial-Style Demos


Try It

Try out some visual analytics of various topic models at

D-VITA now also offers a user model and a control panel for managing and invoking the workflows and tools for obtaining the topic models, which had to be done manually previously. These topic modeling workflows are now supported fully automatically. Also the newest D-VITA build offers some fine architectural decoupling of components under the hood. Unfortunately we cannot offer all new features to the public and there are several improvements in the pipe to make to make handling of topic model requests by external users more efficient.

Get Involved

If you like to get involved in coding for future milestones, please check out the D-VITA repository on Github, and its associated submodules:

If you're interested in the research perspective and want to work on D-VITA as part of your Bachelor thesis, Master thesis, or HiWi job, please contact Dr. Michael Derntl.

Other Media

Below you can find a presentation of an analysis of the Learning Analytics & Knowledge Conference dataset for the LAK Data Challenge 2013 [4]


  1. N. Günnemann, M. Derntl, R. Klamma, M. Jarke: An Interactive System for Visual Analytics of Dynamic Topic Models. Datenbank-Spektrum 13(3): 213-223 (2013)
  2. D. Blei, J. D. Lafferty: Dynamic topic models. Proceedings of ICML 2006. New York: ACM (2006)
  3. M. Derntl, N. Günnemann, A. Tillmann, R. Klamma, M. Jarke: Building and Exploring Dynamic Topic Models on the Web. Proceedings of CIKM 2014. New York: ACM (2014)
  4. M. Derntl, N. Günnemann, R. Klamma: A Dynamic Topic Model of Learning Analytics Research. LAK Data Challenge (2013)

Document Actions