Image Image Image Image Image Image Image Image Image Image

| July 26, 2017

Scroll to top


App Store Topic Models: Visualized

App Store Topic Models: Visualized

| On 25, Apr 2016

In the context of finding the optimal similarity measurement technique of apps in the app store (Think: solving the app discoverability problem), I have been toying around with several typical measurement methods. Among the ones I’ve investigated, is the ever popular topic modelling (using Latent Dirichlet Allocation). As app stores provide short and sweet textual descriptions, we think that these descriptions must account to something, right? A quick beautiful LDA code can reveal really interesting topics that clump apps together based on how they choose to describe themselves and what to advertise in their descriptions. Read more about our research in the UCLappA website.

I have used the Playdrone dataset provided by Nicholas Viennot et al. from the University of Columbia. LDA has been carried out using typical R packages.

Instead of showing you samples of terms and whole bunch of numbers, here is for your viewing pleasure, the topics and their terms visualised using Carson Sievert beautiful LDAVis R Package. [Github]

Open the topic models in a new window. (Looks better, I promise)
Hover on a topic (circle) to see to the right the terms associated with it. Hover on a term to the right to see the topics that include it.