|
Repo | Description | Status | Note |
---|---|---|---|
BotEval |
Completeā |
||
Distilling strong reference based metrics into stronger reference-less metrics |
Completeā |
||
A streaming approach to machine translation training |
Completeā |
||
Many-to-English (v2) |
Completeā |
||
Transformer ablation, showing that model can work without encoder. |
Completeā |
||
Parallel sentence alignment from Universal Declaration of Human Rights corpus |
WIP/Incompleteā |
||
Done |
Completeā |
||
Macro sampling in BERT |
Didnāt workā |
Maybe we should revisit |
|
Imbalanced machine learning: case studies in image recognition, text classification, and machine translation |
Incompleteā |
||
A theory on hyperparameter |
Incompleteā |
Book idea! Needs more time. š |
|
A survey of NMT toolkits |
Incompleteā |
Lost interest |
|
Macro-averaged evaluation for automatic speech recognition |
Incompleteā |
(Some positive results, but needs more evidence) |
|
Macro Average: Rare Types are Important Too |
Completeā |
||
Many-to-English machine translation tools, data, and pretrained models |
Completeā |
||
Finding the optimal vocabulary size for neural machine translation |
Complete ā |
||
Neural machine translation with imbalanced classes |
Complete ā |
Rejected from *ACL; Arxiv link |
|
NMT learning curve revisited. |
Complete ā |
Not published |
|
An Approach for Automatic and Large Scale Image Forensics |
Complete ā |
||
Clustering webpages based on structure and style similarity |
Completeā |