Vis2 Project

Based on

D. Fried and S. G. Kobourov, "Maps of Computer Science,"

The original paper: Maps of Computer Science.

Abstract: A practical approach for visual exploration of research papers. Specifically, we use the titles of papers from the DBLP database to create what we call maps of computer science (MoCS)

Our contributions

Base map

✓ Represent word similarity with actual distance (Word2Vec)
✓ Make the map ”zoomable” (Next.js)
✓ Reduce clutter by loading words only once they are a certain size (Next.js)
✓ Use UMAP’s features instead of Collision detection
✓ Split data into Groups based on TF-IDF

Heat Map

✓ Make it represent “actual” semantic meaning (UMAP + TF-IDF bins)
✓ The Real Time Part: Make the TF-IDF applicable on new Documents (Real Time Computation with FastAPI)
✓ Use pretrained Word2Vec model instead of training for each input (also better results)

Our pipeline

The new general pipeline is following:

Text corpus -> POS Filtering -> TFIDF calculation and filtering -> Embed to vector space by Word2Vec -> UMAP.

How to run

In one console open:

Go to rtvis/python

py main.py

In second console:

Go to rtvis

npm install

npm run dev

Maps of Computer Science

Authors