A website that lists the gists of research articles and categorizes them by topics.

I'm often interested in knowing the precise statement of all the main claim(s) of a research paper, but doing so involves finding the paper's PDF and then manually searching it. This takes time, especially if the paper uses non-standard notation. Ideally, the abstract should give me this information, but it either doesn't, or the useful info is buried in a long paragraph. Hence, I decided that whenever I read a paper (even if I read it briefly), I'm going to note down the main claims. This is useful if I need to look at the paper many times in the future, e.g., when writing the 'Related Work' section of my papers, or when I need to look at multiple papers together to get a high-level overview of the state of research in a topic.

Information about each article is placed in a JSON file. The createWebsite.py script takes as input a directory containing these JSON files and outputs a website. The articles directory contains JSON files that I curated. Example invocation:

python3 createWebsite.py -i articles -o output

You can filter articles based on topics and subtopics. There's currently no UI for this. You'll have to open your browser's console and call the filterArticles function. Here are a few examples:

filterArticles(hasTopicF('marketEquilibrium'));
filterArticles(hasSubtopicF('fairDivision', 'ef1'));

You can compose conditions using functions andF and orF:

filterArticles(andF([hasSubtopicF('fairDivision', 'ef1'), hasSubtopicF('fairDivision', 'prop')]));
filterArticles(orF([hasSubtopicF('fairDivision', 'ef1'), hasSubtopicF('fairDivision', 'prop')]));

filterArticles works by reading articles.json, which is written to the output directory by createWebsite.py. You can also create your own filter functions if you want to filter on something other than topics and subtopics. See script.js to get started.

If you have ideas for a better UI for filtering, please let me know by opening an issue or pull request or contacting me personally.

Python code has been type-annotated. To type-check using mypy, run

mypy --strict createWebsite.py

© 2023 Eklavya Sharma

The JSON files in the articles directory are under CC0. Other than that, all code is licensed under GNU GPLv3. This roughly means that you are free to use, modify, and distribute this data and code.