IArxiv beta
IArxiv.org fetches Arxiv.org papers every day and sorts them according to each user preferences. Therefore, everyday each user can read the Arxiv papers through IArxiv in a much more easy and pleasant way!
There are two ways to learn which paper the user likes. The first one occurs during registration, where IArxiv fetches previous user's papers (based on the user first and last names). And the second occurs everyday when the user clicks on papers to open them.
In both cases IArxiv runs an unsupervised Machine Learning algorithm (called LDA) to extract which topics and in which proportions are present in each paper, to construct a vector of topics for each user. This personal vector of topics is the key to sort new Arxiv releases according to each user preferences.
Usually, scientists enter everyday to check each new Arxiv release. However, if you go in vacations, or you had a busy week, IArxiv also offers to select a slice of dates and sort them altogether for a more easy return to work!
By now astro-ph, gr-qc, hep-ph and hep-th. We plan to include more as time goes by.
Yes, you can choose that in settings. Every time you click a paper, the algorithm learns whether is from one or other category or cross-listing, and learns within the topics of each category.
Yes, and currently working!"