Section: Software
Information Retrieval
CBR*Tools for Managing and Reusing Past Experiences based on Historical Data
Participant : Brigitte Trousse [correspondant] .
CBR*Tools [92] , [93] is an object-oriented framework [94] , [91] for Case-Based Reasoning which is specified with the UMT notation (Rational Rose) and written in Java. It offers a set of abstract classes to model the main concepts necessary to develop applications integrating case-based reasoning techniques: case, case base, index, measurements of similarity, reasoning control. It also offers a set of concrete classes which implements many traditional methods (closest neighbors indexing, Kd-tree indexing, neuronal approach based indexing, standards similarities measurements). CBR*Tools currently contains more than 240 classes divided in two main categories: the core package for basic functionality and the time package for the specific management of the behavioral situations. The programming of a new application is done by specialization of existing classes, objects aggregation or by using the parameters of the existing classes.
CBR*Tools addresses application fields where the re-use of cases indexed by behavioral situations is required. The CBR*Tools framework was evaluated via the design and the implementation of several applications such as Broadway-Web, Educaid, BeCKB, Broadway-Predict, e-behaviour and Be-TRIP.
CBR*Tools is concerned by two past contracts: EPIA and MobiVIP.
CBR*Tools will be available for research, teaching and academic purpose via the FocusLab platform. The user manual can be downloaded at the URL: http://www-sop.inria.fr/axis/cbrtools/manual/ .
See also the web page http://www-sop.inria.fr/axis/cbrtools/manual/ .
Broadway*Tools for Building Recommender Systems on the Web
Participant : Brigitte Trousse [correspondant] .
Broadway*Tools is a toolbox supporting the creation of adaptive recommendation systems on the Web or in a Internet/Intranet information system. The toolbox offers different servers, including a server that computes recommendations based on the observation of the user sessions and on the re-use of user groups' former sessions. A recommender system created with Broadway*tools observes navigations of various users and gather evaluations and annotations, to draw up a list of relevant recommendations (Web documents, keywords, etc).
Based on Jaczynski'thesis [92] , different recommender systems have been developed for supporting Web browsing, but also browsing inside a Web-based information system or for query formulation in the context of a meta search engine.