In this lecture, we will cover training protocols that involve structured inference as a post-processing step. To do so, we will see three general families of approaches for writing down loss functions that involve an inference component: structured variants of SVMs and perceptrons, and conditional random fields.
Lecture slides
Readings
-
Chapter 3 of Noah Smith, Linguistic Structure Prediction.
-
Ioannis Tsochantaridis, Thomas Hofmann, Thorsten Joachims, and Yasemin Altun. Support vector machine learning for interdependent and structured output spaces, ICML 2004.
-
Ben Taskar, Carlos Guestrin, and Daphne Koller. Max-margin Markov networks, NIPS 2004.
-
John Lafferty, Andrew McCallum and Fernando Pereira, Conditional Random Fields: Probabilistic Models for Segmenting and Labeling Sequence Data, ICML 2001.
-
Michael Collins, Discriminative Training for Hidden Markov Models: Theory and Experiments with Perceptron Algorithms, EMNLP 2002