If you are interested in technical details ...

Kvasir is the acronym for Knowledge ViA Semantic Information Retrieval, it is also the name of a Scandinavian god in Norse mythology who travels around the world to teach and spread knowledge and is considered extremely wise. At the core, Kvasir implements an LSA-based index and search service. For more details, please read our technical report as blow.

Technical Report

Kvasir will be demonstrated in WWW'15 conference in Florence, Italy, May 2015.

Source code

Kvasir is an open-source project. The source code of its key components are hosted on Github and publicly accessible. We will release the complete source code of Kvasir after our paper gets accepted. We welcome researchers and developers to use our code, hack it and extend it!

A popular and highly optimized python library for indexing and searching approximate k-nearest neighbors in very high dimensional spaces, with over 50,000+ downloads so far!

Stochastic singular value decomposition for huge matrices on Apache Spark. The fat matrices are first thinned by rank-revealing QR estimation for dimension reduction.

Kvasir frontend is currently implemented as a lightweight Chrome extension using Javascript. The frontend connects the backend with a simple RESTful API.

Coming soon ...