0.25
0.5
0.75
1.25
1.5
1.75
2
Nevronsko strojno prevajanje
Published on Nov 27, 2018931 Views
Predavanje ‘Nevronsko strojno prevajanje’ bo prikazalo trenutno stanje in izzive na področju nevronskega prevajanja. Prikazali bomo kratek uvod v nevronske mreže (nelinearnost, vzvratno razširjanje na
Related categories
Chapter list
Neural Machine Translation00:00
Contents01:37
A Short history02:32
Promising NMT toolkits03:42
Linear models04:49
Multiple layers05:36
Non - linearity06:32
Back - propagation training - 106:47
Back - propagation training - 207:30
Problems with GDT08:18
Refinements - 109:04
Computation graphs10:16
Neural language models11:04
Word embeddings12:43
Feed - forward neural language models13:06
Recurrent neural language models13:13
Back - propagation through time14:05
Long Short-Term Memory models14:46
Gated recurrent units15:49
Deep recurrent neural networks16:20
Neural translation models17:07
Alignment model (attention) - 118:16
Alignment model (attention) - 219:19
Attention mechanism20:14
Training21:12
Parallel mini - batches21:45
Beam search - 122:43
Search graph for beam search 23:37
Beam search - 224:04
Refinements - 224:23
Subword units (BPE) - 125:09
Subword units (BPE) - 225:54
Using monolingual data27:03
Back translation27:56
Round trip training28:38
Domain adaptation29:11
Other refinements30:40
Multiple language pairs31:08
Alternate architectures31:55
Self - Attention (Google)32:16
Hierarchical neural attention encoder32:44
Current challenges - 133:33
Current challenges - 234:22
Literature35:24
Slovenian ⇔ English NMT English - 135:49
Slovenian ⇔ English NMT English - 237:32
Slovenian ⇒ English BLEU39:22
English ⇒ Slovenian BLEU39:46
Questions?40:20