Find a Needle in a Haystack - Sorting arXiv Articles by Your Criteria

Over a thousand articles in machine learning are published on arXiv every week. It becomes increasingly difficult to find new relevant articles in such a stream. The purpose of this article is to offer and describe a software that allows one to sort arXiv articles according to users’ criteria. It can be useful in journal clubs or in research to highlight subjectively interesting articles satisfying user-specified constraints like period of publishing, relevant and irrelevant topics and keywords. The described results were tested at weekly journal clubs and lead to a reduction in time costs by several times. Another purpose of the article was to show typical choices faced by the authors in adjusting this software as well as to show the rationales for the chosen default settings. That can help new users to get start easier and faster. The software can be used also in other research areas outside of machine learning. The source code is available at https://github.com/Schakirov/arXiv-metadata.