Generative and Discriminative Models in Statistical Parsing

author: Michael Collins, Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology, MIT
published: March 26, 2010, recorded: December 2009, views: 477

Slides

Slides
0:00	Generative and Discriminative Models in Statistical Parsing
0:14	Generative and Discriminative Models in Statistical Parsing
1:26	Parsing
1:29	Generative and Discriminative Models in Statistical Parsing
1:40	Parsing
2:01	Outline
2:45	Parsing
2:47	Outline
3:16	Discriminative Model 1: SPATTER
5:31	The Label-Bias Problem
8:06	Discriminative Model 2: Lexical Dependencies (C, 1996)
11:31	Results (Discriminative Model 2)
12:16	Generative Models 1, 2: Markov Grammars
12:29	Markov Grammars (continued) (1)
12:31	Markov Grammars (continued) (2)
12:33	Markov Grammars (continued) (3)
12:37	Generative Models 1, 2: Markov Grammars
12:52	Discriminative Model 2: Lexical Dependencies (C, 1996)
12:58	Markov Grammars (continued) (1)
14:11	Markov Grammars (continued) (2)
14:41	Markov Grammars (continued) (3)
14:56	Markov Grammars (continued) (4)
15:01	Markov Grammars (continued) (5)
15:07	Markov Grammars (continued) (6)
15:13	Markov Grammars (continued) (7)
15:16	Markov Grammars (continued) (8)
15:41	Markov Grammars (continued) (1)
15:47	Markov Grammars (continued) (2)
15:52	Markov Grammars (continued) (3)
15:53	Markov Grammars (continued) (4)
15:54	Markov Grammars (continued) (5)
15:55	Markov Grammars (continued) (6)
15:58	Results (Markov Grammars)
16:35	Markov Grammars (continued) (3)
17:57	Results (Markov Grammars)
18:10	Discriminative Model 3: (McDonald et al, 2005)
20:05	Results (Markov Grammars)
20:10	Discriminative Model 3: (McDonald et al, 2005)
20:41	Discriminative Model 4: a TAG-Based Model (1)
20:52	Discriminative Model 4: a TAG-Based Model (2)
21:17	Discriminative Model 4: a TAG-Based Model (3)
21:31	Discriminative Model 4: a TAG-Based Model (4)
21:32	Discriminative Model 4: a TAG-Based Model (5)
21:34	Discriminative Model 4: a TAG-Based Model (6)
21:40	Discriminative Model 4: a TAG-Based Model (7)
21:45	Discriminative Model 4: a TAG-Based Model (8)
21:55	Markov Grammars (continued) (4)
22:01	Discriminative Model 4: a TAG-Based Model (2)
22:07	Discriminative Model 3: (McDonald et al, 2005)
22:13	Discriminative Model 4: a TAG-Based Model (8)
22:15	Discriminative Model 4: a TAG-Based Model, Trigram dependency features
22:27	Discriminative Model 4: a TAG-Based Model, More trigram dependency features
22:40	Results (Discriminative Model (4)
23:45	“Hybrid” Discriminative/Generative Model 1: Word Clusters
24:55	Results, Dependency accuracy for a 2nd order parser
25:34	“Hybrid” Discriminative/Generative Model 2
26:44	The Generative Models
27:26	Results (“Hybrid”)
28:09	Final Thoughts

Report a problem or upload files

If you have found a problem with this lecture or would like to send us extra material, articles, exercises, etc., please use our ticket system to describe your request and upload the data.
Enter your e-mail into the 'Cc' field, and we will keep you updated with your request's status.

Lecture popularity: You need to login to cast your vote.

Description

Since the earliest work on statistical parsing, a constant theme has been the development of discriminative and generative models with complementary strengths. In this work I’ll give a brief history of discriminative and generative models in statistical parsing, focusing on strengths and weaknesses of the various models. I’ll start with early work on discriminative history-based models (in particular, the SPATTER parser), moving through early discriminative and generative models based on lexicalized (dependency) representations, through to recent work on conditional-random-field based models. Finally, I’ll describe research on semi-supervised approaches that combine discriminative and generative models.