en
0.25
0.5
0.75
1.25
1.5
1.75
2
Never Ending Language Learning
Published on Jul 13, 201212734 Views
Related categories
Chapter list
Never Ending Language Learning00:00
Tenet 101:05
Tenet 202:36
NELL: Never-Ending Language Learner03:39
NELL today06:14
NELL knowledge fragment06:15
NELL today example07:39
Semi-Supervised Bootstrap Learning14:14
Key Idea 1: Coupled semi-supervised training of many functions16:33
Type 1 Coupling: Co-Training, Multi-View Learning18:09
Type 2 Coupling: Multi-task, Structured Outputs20:16
Multi-view, Multi-Task Coupling20:50
Learning Relations between NP’s (1)20:51
Learning Relations between NP’s (2)20:52
Type 3 Coupling: Argument Types20:53
Basic NELL Architecture21:22
NELL: Learned reading strategies21:24
If coupled learning is the key, how can we get new coupling constraints?25:46
Key Idea 2: Discover New Coupling Constraints26:03
Example Learned Horn Clauses27:17
Some rejected learned rules27:21
Learned Probabilistic Horn Clause Rules27:25
Key Idea 3: Automatically extend ontology29:05
Ontology Extension (1)30:06
Example Discovered Relations31:16
NELL: recently self-added relations32:35
Key Idea 4: Cumulative, Staged Learning33:25
Learning to Reason at Scale36:14
Inference by KB Random Walks (1)37:12
Inference by KB Random Walks (2)39:16
CityLocatedInCountry(Pittsburgh) = ? (1)40:07
CityLocatedInCountry(Pittsburgh) = ? (2)40:31
CityLocatedInCountry(Pittsburgh) = ? (3)40:43
CityLocatedInCountry(Pittsburgh) = ? (4)40:50
CityLocatedInCountry(Pittsburgh) = ? (5)41:20
CityLocatedInCountry(Pittsburgh) = ? (6)41:23
CityLocatedInCountry(Pittsburgh) = ? (7)41:28
CityLocatedInCountry(Pittsburgh) = ? (8)41:32
CityLocatedInCountry(Pittsburgh) = ? (9)41:33
CityLocatedInCountry(Pittsburgh) = ? (10)42:53
Random walk inference: learned path types42:56
Random walk inference: example (1)43:29
Random walk inference: example (2)44:02
Summary45:59
Thank you46:05