MeanFilter aims to provide an efficient and smart way to identify threats from internet posts. Using the power of Nuance, we built training sets to distinguish an author's intent. Currently the application only supports Twitter but is abstract enough to scale to multiple platforms (Facebook, Reddit, etc). A possible extent of our bot would be to manage a Celebrity Facebook page to report or delete undesired content.


MeanFilter is a combination of multiple technologies.

  • Java for the main processing
  • NodeJs to interface with Nuance's API
  • AngularJS for the user interface

Each one of the services communicate with each others using web sockets to achieve optimal performance.


It took us a lot of time to figure out that Nuance's API only supported W3C compliant web sockets. The HTTP upgrade request was mentioning a certain RFC although it did not fully support it. We were pretty happy to overcome this difficulty and gather a deeper knowledge of this technology.

Share this project: