Knowledge Pump provides users with personalized recommendations for things to read. When users sign up, they join communities of people with similar interests. Profiler agents track and map each user's interests, learning more about the person each time (s)he uses the Pump. A recommender agent finds matches between new items and user preferences, automatically sending relevant and high quality information to people as it is found.
Knowledge Pump consists of a set of agents providing on-line support for existing intranet- and extranet-based communities. The Knowledge Pump channels the flow and use of knowledge in an organization, connecting document repositories, people and processes, and leveraging formal and informal organizational charts and structure. In particular, the main objective of the Knowledge Pump is to help communities, defined by their common interests and practices, more effectively and more efficiently share knowledge, be it in the form of must-read documents or new ways to get work done.
The core of the Knowledge Pump is the recommendation functionality that is based on community-centred collaborative filtering which filters both by content and by taste. The Pump handles content filtering by relying on recommenders to classify items into pre-defined communities. Social filtering matching items to people by first matching people to each other is accomplished using a combination of statistical algorithms and boot-strapped snapshots of the underlying social network of a collection of users.
More like this
- 2016/112 - Enriching How-to Guides with Actionable Phrases and Linked Data
- Co-construction of adaptive public policies using SmartGov
- Moment Matching Training for Neural Machine Translation: A Preliminary Study
- Research scientist - Search and Recommendation
- American Association for Applied Linguistics (AAAL) 2018, Chicago, USA