Class Meeting for 10-710 09-01-2011

This is one of the class meetings on the schedule for the course Structured Prediction 10-710 in Fall 2011.

Hidden Markov Models as Structured Prediction

• Slides

Required Readings

• Borkar 2001 Automatic Segmentation of Text Into Structured Records
• Frietag 2000 Maximum Entropy Markov Models for Information Extraction and Segmentation

Optional Readings

• An Algorithm that Learns What's in a Name, Bikel et al, MLJ 1999. Another well-engineered and influential HMM-based NER system.
• Unsupervised Learning of Field Segmentation Models for Information Extraction, Grenager, Klein, and Manning, ACL 2005. Unsupervised segmentation paper.
• Named Entity Recognition with Character-Level Models, Klein et al, CoNLL 2003. Interesting twist on the standard approach of token-tagging - NER by tagging characters and character n-grams.

Background Readings

• LSP: chapter 1 (through 1.3) and section 3.3 (through 3.3.3); for some background on dynamic programming (buildup to Viterbi algorithm), see 2.3 (through 2.3.1, beyond if you're interested)
• A Maximum Entropy Part-Of-Speech Tagger, Ratnaparkhi, Workshop on Very Large Corpora 1996
• Mike Collins on learning in NLP, including a section on maxent taggers.
• Dan Klein on maxent.
• Bob Carpenter on maxent and SGD has detailed derivations of multiclass logistic regression and its gradients.