Machine Learning and Cognitive Science of Language Acquisition

Unsupervised Learning of Syntactic Structure

author: Chris Manning, Departments of Computer Science and Linguistics, Stanford University

Description

Probabilistic models of language. ''Everybody knows that language is variable'' - Sapir (1921).
Probabilistic models give precise descriptions of a variable, uncertain world. The choice for language isn’t a dichotomy between rules and neural networks. Probabilistic models can be used over rich linguistic representations. They support inference and learning. There’s not much evidence of a poverty of the stimulus preventing them being used.

You might be experiencing some problems with Your Video player.

*Slides*
0:00	Unsupervised Learning of Syntactic Structure
0:02	Probabilistic models of language
1:27	Poverty of the stimulus?
2:27	Linguistics:Is it time for a change?
3:24	Gold (1967)
6:04	Horning (1969)
9:53	1. Grammar induction:Learning phrase structure
10:11	Idea: Lexical Affinity Models
11:42	Problem: Non-Syntactic Affinity
13:19	Idea: Word Classes
13:43	Problems: Word Class Models
15:03	Bias: Using more sophisticated dependency representations
16:29	Constituency Parsing
17:54	Idea: Learn PCFGs with EM
18:20	Other Approaches
18:43	Right-Branching Baseline
19:11	Inspiration: Distributional Clustering
20:09	Idea: Distributional Syntax?
22:41	Problem: Identifying Constituents
24:13	A Nested Distributional Model
25:04	Constituent-Context Model (CCM)
26:42	Initialization: A little UG?
28:44	Results: Constituency
29:31	A combination model
29:41	Combining the two models [Klein and Manning ACL 2004]
30:38	Crosslinguistic applicability of the learning algorithm
30:59	Most Common Errors: English
31:03	What Has Been Accomplished?
31:05	2. Unsupervised learning of linking to semantic roles
32:04	Semantic Role Labeling
32:48	Why Unsupervised Might Work - part 1
33:09	Why Unsupervised Might Work - part 2
33:46	Inputs and Outputs
34:09	Probabilistic Model
35:59	Linking Model
36:35	Linking Construction Example
37:21	Learning Problem
38:16	Datasets and SRL Evaluation
41:02	Results (Coarse Roles, Sec. 23)
42:22	Improved Verbs
44:16	Conclusions

Lecture rating

People found this lecture:
Worth seeing
because it is:
Valuable and informative
Well presented
Easily understandable
Acceptably recorded
You need to login to cast your vote.

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.