Speaker: Ariadna Quattoni, LSI<br />
Title: &quot;Tutorial on Conditional Random Fields for Sequence Prediction &quot;<br />
Time: Tuesday November 24th, 15:30<br />
Place: Room S208, floor -2, Omega Building<br />
<br />
Abstract:<br />
Many applications in machine learning can be casted as sequence prediction problems, where given a sequence of input symbols the goal is to predict an output sequence that assigns a label to each input symbol. In natural language processing, for example, the problem of part of speech tagging involves predicting a sequence of part of speech tags for a given sentence (i.e. a sequence of words). <br />
<br />
Conditional Random Fields (CRF) have been proposed to solve such sequence prediction problems. A CRF models the distribution of output sequences conditioned on an input sequence as a markov random field. <br />
<br />
In this tutorial I will present the model and training algorithms. I will also frame CRFs as an instance of a larger class of factorized linear models for predicting structures. Finally, I will describe some extensions of the CRF model that allow for the incorporation of hidden variables.