Machine Learning and Cognitive Science of Language Acquisition

A Bayesian approach to Word Segmentation: Theoretical and Experimental results

author: Sharon Goldwater, Department of Linguistics, Stanford University

Description

Word segmentation. One of the first problems infants must solve when learning language. Infants make use of many different cues - phonotactics, allophonic variation, metrical (stress) patterns, effects of coarticulation, and statistical regularities in syllable sequences. Statistics may provide initial bootstrapping - used very early (Thiessen & Saffran, 2003); language-independent.

You might be experiencing some problems with Your Video player.

*Slides*
0:00	A Bayesian approach to word segmentation: Theoretical and experimental results
0:15	Word segmentation
1:16	Modeling statistical segmentation
1:53	Outline
2:31	Statistical segmentation
3:25	Interpretation of TPs
4:14	Questions
5:04	Bayesian learning
5:55	Bayesian segmentation
6:58	Brent (1999)
7:43	A new unigram model (Dirichlet process) - part 1
8:31	A new unigram model (Dirichlet process) - part 2
9:50	Unigram model: simulations
10:35	Example results
10:57	Comparison to previous results
11:25	What happened?
12:05	What about other unigram models?
13:06	Bigram model (hierachical Dirichlet process)
13:42	Example results
13:55	Quantitative evaluation
14:25	Summary
15:12	Remaining questions
15:46	Testing model predictions
16:22	Experimental method
17:17	Procedure
17:58	Human performance
18:24	Model comparison
18:57	Models used
19:51	Results: linear fit
21:06	Results: words vs. part-words
22:30	Summary
23:24	Continuing work
24:25	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.