A Topic Hex Map is a tool we developed to summarize any text portfolio in an structured hexagonal grid. Using only text as input, we extract the topics from it and place them as hexagons on a regular grid. This map format allows for quick and efficient browsing of the core subjects in the corpus supplied.
.png)
Extracting the topics from a collection of document is done without any supervision, using Latent Dirichlet Allocation1 (LDA). The only inputs necessary are the texts and the number of topics required. After the topic modelling process, we perform an additional algorithmic step to estimate similarities between topics. The final mapping process then uses that similarity information to place the topics on a grid.
1 Blei, D.M., Ng, A.Y. and Jordan, M.I., 2003. Latent dirichlet allocation. Journal of machine Learning research, 3(Jan), pp.993-1022.
9 Years of UK & EU Research.
Inspired by Brexit, we used Topic Hex Maps to compare British and European funded grants.