index.md

relevanced is a multithreaded C++ server providing efficient document similarity scoring and classification as a networked service.

Using the established vector space model, it helps you answer questions like:

• How similar are these two documents?
• How similar is this new document to a collection of other documents?
• Which of several groups of documents is most like this new document?

relevanced can process text in English, Italian, German, French, Spanish and Russian. It is unicode-aware and uses UTF-8 for all internal string representations.

It uses an out-of-core model retraining approach to handle massive document collections with limited memory.

Clients are available for Python, Java, Javascript and Ruby.

The server can be installed from a DEB package or run as a Docker image, or you can build it from source.

Released under the commercial-friendly MIT license.