Learning from Partially Annotated Sequences thumbnail
Pause
Mute
Subtitles
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

Learning from Partially Annotated Sequences

Published on Nov 30, 20112651 Views

We study sequential prediction models in cases where only fragments of the sequences are annotated with the ground-truth. The task does not match the standard semi-supervised setting and is highly rel

Related categories

Chapter list

Protein Secondary Structure Prediction00:00
Part-of-Speech Tagging00:24
Named Entity Recognition01:16
Representation01:40
Structural Perceptrons03:11
Viterbi Algorithm (only transitions) - 104:22
Viterbi Algorithm (only transitions) - 204:44
Labeled Data is Scarce05:25
Research Question06:58
A Simple Transductive Perceptron07:48
Constrained Viterbi Algorithm - 109:08
Constrained Viterbi Algorithm - 209:23
Transductive Loss-augmented Perceptron09:42
Experiments10:22
Varying % of labeled tokens11:30
Adding Partially Labeled Ex.12:17
Automatically + Partially Labeled Data12:51
Wikipedia Corpus - 114:19
Wikipedia Corpus - 214:31
Wikipedia Corpus - 314:39
Wikipedia Corpus - 414:51
Wikipedia Corpus - 515:13
Wikipedia Corpus - 615:16
Characteristics (English)16:03
Mono-lingual Experiment17:06
Cross-lingual NER18:15
Cross-lingual Label Propagation - 118:59
Cross-lingual Label Propagation - 219:03
Cross-lingual Corpus English → Spanish19:11
Characteristics (Spanish)19:20
Cross-lingual Experiment19:54
Conclusion20:19