add sentence & character level data augmentation api

[lugimzzz]

PR types

New features

PR changes

APIs

Description

新增句子和字级别数据增强策略，新增单词级别词表。

• 句子：相似句生成、回译、句子摘要、句子续写
• 字：替换、删除、插入、交换
• 词：新增基于word-embedding的近义词词表、反义词词表
• 文档：新增文档输入增强。

[paddle-bot]

Thanks for your contribution!

[codecov]

Codecov Report

Merging #4194 (1b106e2) into develop (0b1d706) will increase coverage by 1.44%.
The diff coverage is 67.96%.

@@             Coverage Diff             @@
##           develop    #4194      +/-   ##
===========================================
+ Coverage    40.10%   41.54%   +1.44%     
===========================================
  Files          439      438       -1     
  Lines        61568    62142     +574     
===========================================
+ Hits         24689    25816    +1127     
+ Misses       36879    36326     -553     

Impacted Files	Coverage Δ
paddlenlp/dataaug/base_augment.py	63.30% <21.95%> (+35.67%)	⬆️
paddlenlp/dataaug/word.py	67.24% <67.24%> (ø)
paddlenlp/dataaug/sentence.py	68.52% <68.52%> (ø)
paddlenlp/dataaug/char.py	73.38% <73.38%> (ø)
paddlenlp/dataaug/__init__.py	100.00% <100.00%> (ø)
paddlenlp/taskflow/task.py	35.90% <0.00%> (-0.43%)	⬇️
paddlenlp/transformers/roberta/modeling.py	89.81% <0.00%> (-0.37%)	⬇️
paddlenlp/transformers/gpt/modeling.py	78.42% <0.00%> (+0.20%)	⬆️
...lenlp/ops/fast_transformer/transformer/decoding.py	7.72% <0.00%> (+0.22%)	⬆️
paddlenlp/transformers/albert/modeling.py	84.63% <0.00%> (+0.23%)	⬆️
... and 21 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[sijunhe]

这里需要把tests/dataaug加入https://github.com/PaddlePaddle/PaddleNLP/blob/develop/pyproject.toml#L15 ，单测才能跑起来

[wawltor]

2022->2023

[lugimzzz]

已修改

[wawltor]

2022->2023

[lugimzzz]

同上

[lugimzzz]

这里需要把tests/dataaug加入https://github.com/PaddlePaddle/PaddleNLP/blob/develop/pyproject.toml#L15 ，单测才能跑起来

已添加

[wawltor]

LGTM