Activity: Talk or presentation › Lecture and oral contribution
We present our experiences in using probabilistic-logic models in bioinformatics, computational linguistics and other modeling and prediction tasks.We start giving an introduction to PRISM models that has been developed by T.Sato and his colleagues in Tokyo; this is basically Prolog extended with random variables and supported with machine learning and generalized viterbi calculations.Examples of such models are given and we discuss our current research topics: optimizing such models so they may run in reasonable time, how we may combine several models into one (coordinating models) and our considerations on extending them with constraints for enhanced expressibility.We discuss also what this sort of models offer to computational biology that is not possible with existing standards such standard HMMs and SCFGs.This work is part a larger, funded project, http://lost.ruc.dk, funded by the Danish Strategic Research Council.
17 Apr 2009
Probabilistic-logic models for analysis of language and biological sequence data
Universitat Rovira i Virgili, Campus Catalunya, Departament de Filologies Romàniques