Arxiv Viewer is a productivity chrome extension tool enabling quicker understanding of cited works on arXiv research papers. It's built using the amazing open source web PDF viewer, PDF.js.
The best way to install is from the Chrome Web Store directly here.
You can also install from source with the following commands. First clone this repository.
git clone https://github.com/tchittesh/arxiv-viewer.git
cd arxiv-viewer
Then, install Node.js via the official package or via nvm. Finally, install gulp, other dependencies, and build the chrome extension:
npm install -g gulp-cli
npm install
gulp chromium
Finally, open Chrome, go to More Tools > Extension and load the (unpackaged) extension from the directory build/chromium.
Every LaTeX+BibTeX compiled PDF document handles references to the cited works using internal hyperlinks. These point to a physical location in the pdf (page number, x, y coordinates), and they also contain a unique descriptor d of the cited work that is defined in the bibliography as something like
@article{d,
...
}
and referenced in the LaTeX source as \cite{d}. The current solution is to use this unique descriptor d, as it's by default of the form [last name of first author][year published][first word of title]. Once we extract these three fields, we make a call to arXiv's public API, and display a popup if we generate nonempty nonambiguous results.
Please let me know if you have any suggestions on how to improve the search process. Using Google Scholar would be optimal, as it contains a superset of arXiv papers and allows explicit searching by year; however, they do not have a public API. It would also be great to actually parse the bibliography at the physical location stored in the hyperlink; however, this seems nontrivial.