Sequence Labeling, Contd Prof. Sameer Singh CS 295: STATISTICAL NLP WINTER 2017 February 2, 2017 Based on slides from Dan Klein, and everyone else he copied from.
Outline Marginal Inference in HMMs: F/B algorithm Maximum Entropy Markov Models Conditional Random Fields Neural Sequence Tagging CS 295: STATISTICAL NLP (WINTER 2017) 2
Outline Marginal Inference in HMMs: F/B algorithm Maximum Entropy Markov Models Conditional Random Fields Neural Sequence Tagging CS 295: STATISTICAL NLP (WINTER 2017) 3
Expectation Maximization K-Means Initialization Pick K random centroids Label Data from the Model Cluster all the points Update the Model from Data Update centroids CS 295: STATISTICAL NLP (WINTER 2017) 4
Label Data from the Model Hard-EM Soft-EM CS 295: STATISTICAL NLP (WINTER 2017) 5
Dynamic Programming CS 295: STATISTICAL NLP (WINTER 2017) 6
Forward Backward Algorithm CS 295: STATISTICAL NLP (WINTER 2017) 7
Updating the Model from Data Hard-EM Soft-EM CS 295: STATISTICAL NLP (WINTER 2017) 8
Outline Marginal Inference in HMMs: F/B algorithm Maximum Entropy Markov Models Conditional Random Fields Neural Sequence Tagging CS 295: STATISTICAL NLP (WINTER 2017) 9
Named Entity Recognition George W. Bush spoke from the White House today . PER PER PER O O O LOC LOC O O B-PER I-PER I-PER O O O B-LOC I-LOC O O B-PER I-PER E-PER O O O B-LOC E-LOC O O CS 295: STATISTICAL NLP (WINTER 2017) 10
Max. Entropy Markov Models CS 295: STATISTICAL NLP (WINTER 2017) 11
Graphical Model Notation S E HMMs S E MEMMs CS 295: STATISTICAL NLP (WINTER 2017) 12
Adding Features (for POS) Current Word Window Words CS 295: STATISTICAL NLP (WINTER 2017) 13
Adding Features S E CS 295: STATISTICAL NLP (WINTER 2017) 14
Predictions Using MEMMs Greedy Viterbi Decoding CS 295: STATISTICAL NLP (WINTER 2017) 15
Training MEMMs CS 295: STATISTICAL NLP (WINTER 2017) 16
Outline Marginal Inference in HMMs: F/B algorithm Maximum Entropy Markov Models Conditional Random Fields Neural Sequence Tagging CS 295: STATISTICAL NLP (WINTER 2017) 17
Label Bias Problem B B B S I I I E O O O B B CS 295: STATISTICAL NLP (WINTER 2017) 18
Conditional Random Fields CS 295: STATISTICAL NLP (WINTER 2017) 19
Graphical Model Notation S E HMMs/MEMMs S E CRFs CS 295: STATISTICAL NLP (WINTER 2017) 20
Predictions Using CRFs CS 295: STATISTICAL NLP (WINTER 2017) 21
Likelihood Training of CRFs CS 295: STATISTICAL NLP (WINTER 2017) 22
Likelihood Training of CRFs CS 295: STATISTICAL NLP (WINTER 2017) 23
Forward-Backward Algorithm CS 295: STATISTICAL NLP (WINTER 2017) 24
Outline Marginal Inference in HMMs: F/B algorithm Maximum Entropy Markov Models Conditional Random Fields Neural Sequence Tagging CS 295: STATISTICAL NLP (WINTER 2017) 25
Simple Neural Tagger NN PRP VB food Start I love End CS 295: STATISTICAL NLP (WINTER 2017) 26
MEMM -ish Neural Tagger E NN S PRP VB food Start I love End CS 295: STATISTICAL NLP (WINTER 2017) 27
Recurrent Neural Tagger NN PRP VB food I love CS 295: STATISTICAL NLP (WINTER 2017) 28
Bidirectional RNN Tagger NN PRP VB I love food CS 295: STATISTICAL NLP (WINTER 2017) 29
Upcoming… Homework 2 is due (~10 days): February 13, 2017 • Homework Write-up, data, and code for Homework 2 is up • Ask questions early! • Proposal is due on Tuesday: February 7, 2017 • Project Only 2 pages • Paper summaries: February 17, February 28, March 14 • Summaries Only 1 page each • CS 295: STATISTICAL NLP (WINTER 2017) 30
Recommend
More recommend