information = chaos;
UniversalRate or show the quality of any (web)-document or message.
XMLQuery or submit ratings via http-XML protocol.
Collaborative FilteringThe recommender module gives you tips from people with similar tastes to you.
(Anti)-BookmarkingStore your favorites with RATED.IT and define anti-bookmarks, which help you to avoid seeing sites that are similar to ones you know you dislike.
Open SourceIntegrate rating & filtering functionality into your application using SELECT open source modules (proxy, CGI).
Own or hosted serverSet up your own SELECT server or define your own rating categories at our production server.
GroupwareWe integrate rating & filtering mechanisms with groupware platforms, such as net.board or KOM2000.
Proxy and MacrosLet SELECT inject ratings into all pages you surf using a local or a public proxy service or a set of macros.
Starting with a Specification Document, which gives a general overview on the data architecture, a set of tools have been implemented and tested. A Technical Documentation (800k) explains the object oriented approach and how this design is mapped into a database structure for persistent storage of the manipulated objects. The Advanced Technical Documentation (600kB) gives information e.g. on the Collaborative Filtering Agent Architecture, how to set up a SELECT server and it includes user manuals e.g. for the GroupMark feature. A set of front ends are available, such as cgi, bookmarklets, or proxy. For those who want to build their own rating front-ends, please refer to the List of Protocols and soon we will make an Open Source Toolkit available.
Rob Procter's paper about SELECT: Social and Collaborative Filtering of Web Documents and News
presented at the 5th and 6th User Interfaces for All workshops (1999 and 2000) gives a more detailled overview on
the research done so far and the project itself.
Check out http://select.pvl.at/select_integration.html for a short tutorial about how to integrate select to your web-documents and services.
Introducing virtual money to a recommender service is something we would like to add in a follow-up project.
General rating issues have been worked out in a study in the Web4Groups project (300k zipped doc or rtf).
The collaborative filtering algorithms and test results are described in a paper from Johan Kaers.
A simple and basic implementation of subject tree rating is also available:
Work about a SELECT-compliant Usenet client is documented in: Christopher Lueg (2000), A Flexible and Non-Intrusive Approach to Distribute Context-Specific Information to Usenet News Clients. Accepted at the OzCHI 2000 Conference on Human-Computer Interaction on "Interfacing reality in the new millennium", 4-8 December 2000, Sydney, Australia.
The research project related document collection with complete Javadoc (restricted to project members):
A radio show on recommender services can be listened at Internet on the Air, where related projects are listed as well.
Here follows a list of the partners of the SELECT research consortium: