The atlas is a web-based tool that provides an interactive map of the of research papers from the ArXiv repository. You can navigate categories, discover connections between papers, and papers, and get recommendations based on your interests.

Every paper is connected to the 10 most similar papers based on the similarity of their abstracts.

You can search for papers by title, author, or using a semantic (smart) search. The map will update to show the papers that match your search.

For every paper you can find recommendations for similar papers which may be of interest to you if you liked the paper.

You can also filter by category by clicking on the small swatch in the legend.

The atlas is designed for researchers, students, and anyone interested in exploring the interested in exploring the vast world of scientific research.

The underlying method I use is based on the idea of semantic vector embeddings, a mathematical representation of words and of words and documents in a high-dimensional space.

I generate these embeddings using a BERT-type language model to to capture the underlying semantic meaning of the documents in a vector. This allows me to use the power of vector operations to compute the to compute the similarity between papers.

This similarity is then used to create a graph of papers in which similar papers are connected and grouped close together.

For more information on the method, please refer to the Github Repo.

Since the recommendations are based solely on the abstracts of the papers, the similarity may not always reflect the actual content of the paper. If an abstract is not representative of the paper, the final map will not be accurate.

Also, I limited the scope to the ArXiv repository, which may not always contain all papers, and papers may not be completely up to date.