This repo aims to build on the work of Cornell jsLDA and pyLDAvis by providing some added functionalities to allow for better understanding of topics within a document corpus.
The dashboard is currently not yet deployed or dockerized. For now, the instructions below are for running the dashboard on your local machine
cd osprey_admin
pip3 install -r requirements.txt
python manage.py runserver 8000
4 Open on http://127.0.0.1:8000/ to view the dashboard
a. Select the number of topics to train the LDA model with.
b. Enter in Column to Size, the exact name of the column that we want to size with the LDA algorithm.
c. Enter in Defining Column ID, the column that defines each unique value in Column to Size
d. Upload the CSV and run the model.
Charts are coded in D3.js and Chart.js. Backend calculations are done in DJango. Latent Dirichlet allocation is described in Blei et al. (2003) and Pritchard et al. (2000). Intertopic Distances are calculated as described in LDAvis, using Jenson Shannon Divergence between topic words Endres, D. M.; J. E. Schindelin (2003) and scaling the multi-dimensional data onto 2 dimensions using Principle Component Analysis (PCA) Pearson K (1901)
** Design of the dashboard in the .gif might lag behind the current status of the project and is not reflective of the current state
Upcoming features, design plans and progress can be found on my notion page here