“FreeNeb model training”版本间的差异

来自cslt Wiki
跳转至： 导航、 搜索
（以“==AM== {| class="wikitable" !Task !! Priority || Subtasks !! Deadline !! Status track |- |Japanese AM training with data augmentation || 1 || # Data processing: Add...”为内容创建页面）
 
第3行： 第3行：
 
!Task !! Priority || Subtasks !! Deadline !! Status track
 
!Task !! Priority || Subtasks !! Deadline !! Status track
 
|-
 
|-
−
|Japanese AM training with data augmentation
+
|Japanese 16k AM training with data augmentation
 
||
 
||
 
1
 
1
第12行： 第12行：
 
# Local model training.
 
# Local model training.
 
||
 
||
−
May.16
+
May.18
 
||
 
||
−
 
+
Holding
 
|-
 
|-
 
|-
 
|-
−
|Uyhgur AM training with data augmentation
+
|Chinese 16k children AM training
 +
||
 +
1
 +
||
 +
# Data preparation.
 +
# Online model training.
 +
||
 +
May.31
 +
||
 +
Holding
 +
|-
 +
|-
 +
|Uyhgur 16k AM training with data augmentation
 
||
 
||
 
1
 
1
第24行： 第36行：
 
# Resource file preparation and release.
 
# Resource file preparation and release.
 
# Online model training.
 
# Online model training.
−
# Local model training.
 
 
||
 
||
−
May.16
+
May.25
 
||
 
||
−
 
+
Holding
 
|-
 
|-
  
−
|Chinese AM training with data augmentation||
+
|Chinese 8k AM training||
 +
2
 +
||
 +
# Experiments on small dataset.
 +
# Model training.
 +
||
 +
June.30
 +
||
 +
Holding
 +
|-
 +
|Chinese 16k AM training with data augmentation||
 
3
 
3
 
||
 
||
第39行： 第60行：
 
# Local model training.
 
# Local model training.
 
||
 
||
−
June.9
+
July.10
 
||
 
||
−
 
+
Holding
 +
|-
 +
|Chinese 16k d-vector based AM training||
 +
3
 +
||
 +
# Recipe preparation and experiments on small dataset.
 +
# Model training.
 +
||
 +
June.31
 +
||
 +
Holding
 
|-
 
|-
 
|}
 
|}
第47行： 第78行：
  
 
==LM==
 
==LM==
 +
{| class="wikitable"
 +
!Task !! Priority || Subtasks !! Deadline !! Status track
 +
|-
 +
|Chinese general domain LM training.
 +
||
 +
1
 +
||
 +
# Corpus preparation. @WD
 +
# Vocab list construction. @LR
 +
# LM training. @WY
 +
||
 +
June.30
 +
||
 +
# In progress.
 +
# In progress.
 +
# Waiting for corpus and vocab list.
 +
|-
 +
|Chinese domain specific vacab list construction
 +
||
 +
2
 +
||
 +
# Financial domain keyword construction.
 +
||
 +
July.31
 +
||
 +
Holding.
 +
|-
 +
|}

2018年5月9日 (三) 05:06的版本

AM

Task Priority Subtasks Deadline Status track
Japanese 16k AM training with data augmentation

1

  1. Data processing: Add noise, echo, change speed....
  2. Resource file preparation and release.
  3. Online model training.
  4. Local model training.

May.18

Holding

Chinese 16k children AM training

1

  1. Data preparation.
  2. Online model training.

May.31

Holding

Uyhgur 16k AM training with data augmentation

1

  1. Data processing: Add noise, echo, change speed....
  2. Resource file preparation and release.
  3. Online model training.

May.25

Holding

Chinese 8k AM training

2

  1. Experiments on small dataset.
  2. Model training.

June.30

Holding

Chinese 16k AM training with data augmentation

3

  1. Data processing: Add noise, echo, change speed....
  2. Resource file preparation and release.
  3. Online model training.
  4. Local model training.

July.10

Holding

Chinese 16k d-vector based AM training

3

  1. Recipe preparation and experiments on small dataset.
  2. Model training.

June.31

Holding


LM

Task Priority Subtasks Deadline Status track
Chinese general domain LM training.

1

  1. Corpus preparation. @WD
  2. Vocab list construction. @LR
  3. LM training. @WY

June.30

  1. In progress.
  2. In progress.
  3. Waiting for corpus and vocab list.
Chinese domain specific vacab list construction

2

  1. Financial domain keyword construction.

July.31

Holding.

取自“http://www.cslt.org/mediawiki/index.php?title=FreeNeb_model_training&oldid=30284”