This python library helps you with augmenting nlp for your machine learning projects. Visit this introduction to understand about Data Augmentation in NLP. Augmenter is the basic element of augmentation while Flow is a pipeline to orchestra multi augmenter together.
- Generate synthetic data for improving model performance without manual effort
- Simple, easy-to-use and lightweight library. Augment data in 3 lines of code
- Plug and play to any machine leanring/ neural network frameworks (e.g. scikit-learn, PyTorch, TensorFlow)
- Support textual and audio input
| Section | Description |
|---|---|
| Quick Demo | How to use this library |
| Augmenter | Introduce all available augmentation methods |
| Installation | How to install this library |
| Recent Changes | Latest enhancement |
| Extension Reading | More real life examples or researchs |
| Reference | Reference of external resources such as data or model |
- Quick Example
- Example of Augmentation for Textual Inputs
- Example of Augmentation for Multilingual Textual Inputs
- Example of Augmentation for Spectrogram Inputs
- Example of Augmentation for Audio Inputs
- Example of Orchestra Multiple Augmenters
- Example of Showing Augmentation History
- How to train TF-IDF model
- How to train LAMBADA model
- How to create custom augmentation
- API Documentation
| Augmenter | Target | Augmenter | Action | Description |
|---|---|---|---|---|
| Textual | Character | KeyboardAug | substitute | Simulate keyboard distance error |
| Textual | OcrAug | substitute | Simulate OCR engine error | |
| Textual | RandomAug | insert, substitute, swap, delete | Apply augmentation randomly | |
| Textual | Word | AntonymAug | substitute | Substitute opposite meaning word according to WordNet antonym |
| Textual | ContextualWordEmbsAug | insert, substitute | Feeding surroundings word to BERT, DistilBERT, RoBERTa or XLNet language model to find out the most suitlabe word for augmentation | |
| Textual | RandomWordAug | swap, crop, delete | Apply augmentation randomly | |
| Textual | SpellingAug | substitute | Substitute word according to spelling mistake dictionary | |
| Textual | SplitAug | split | Split one word to two words randomly | |
| Textual | SynonymAug | substitute | Substitute similar word according to WordNet/ PPDB synonym | |
| Textual | TfIdfAug | insert, substitute | Use TF-IDF to find out how word should be augmented | |
| Textual | WordEmbsAug | insert, substitute | Leverage word2vec, GloVe or fasttext embeddings to apply augmentation | |
| Textual | BackTranslationAug | substitute | Leverage two translation models for augmentation | |
| Textual | ReservedAug | substitute | Replace reserved words | |
| Textual | Sentence | ContextualWordEmbsForSentenceAug | insert | Insert sentence according to XLNet, GPT2 or DistilGPT2 prediction |
| Textual | AbstSummAug | substitute | Summarize article by abstractive summarization method | |
| Textual | LambadaAug | substitute | Using language model to generate text and then using classification model to retain high quality results | |
| Signal | Audio | CropAug | delete | Delete audio's segment |
| Signal | LoudnessAug | substitute | Adjust audio's volume | |
| Signal | MaskAug | substitute | Mask audio's segment | |
| Signal | NoiseAug | substitute | Inject noise | |
| Signal | PitchAug | substitute | Adjust audio's pitch | |
| Signal | ShiftAug | substitute | Shift time dimension forward/ backward | |
| Signal | SpeedAug | substitute | Adjust audio's speed | |
| Signal | VtlpAug | substitute | Change vocal tract | |
| Signal | NormalizeAug | substitute | Normalize audio | |
| Signal | PolarityInverseAug | substitute | Swap positive and negative for audio | |
| Signal | Spectrogram | FrequencyMaskingAug | substitute | Set block of values to zero according to frequency dimension |
| Signal | TimeMaskingAug | substitute | Set block of values to zero according to time dimension | |
| Signal | LoudnessAug | substitute | Adjust volume |
| Augmenter | Augmenter | Description |
|---|---|---|
| Pipeline | Sequential | Apply list of augmentation functions sequentially |
| Pipeline | Sometimes | Apply some augmentation functions randomly |
The library supports python 3.5+ in linux and window platform.
To install the library:
pip install numpy requests nlpaugor install the latest version (include BETA features) from github directly
pip install numpy git+https://github.com/makcedward/nlpaug.gitor install over conda
conda install -c makcedward nlpaugIf you use BackTranslationAug, ContextualWordEmbsAug, ContextualWordEmbsForSentenceAug and AbstSummAug, installing the following dependencies as well
pip install torch>=1.6.0 transformers>=4.11.3 sentencepieceIf you use LambadaAug, installing the following dependencies as well
pip install simpletransformers>=0.61.10If you use AntonymAug, SynonymAug, installing the following dependencies as well
pip install nltk>=3.4.5If you use WordEmbsAug (word2vec, glove or fasttext), downloading pre-trained model first and installing the following dependencies as well
from nlpaug.util.file.download import DownloadUtil
DownloadUtil.download_word2vec(dest_dir='.') # Download word2vec model
DownloadUtil.download_glove(model_name='glove.6B', dest_dir='.') # Download GloVe model
DownloadUtil.download_fasttext(model_name='wiki-news-300d-1M', dest_dir='.') # Download fasttext model
pip install gensim>=4.1.2If you use SynonymAug (PPDB), downloading file from the following URI. You may not able to run the augmenter if you get PPDB file from other website
http://paraphrase.org/#/downloadIf you use PitchAug, SpeedAug and VtlpAug, installing the following dependencies as well
pip install librosa>=0.9.1 matplotlib- Return list of output
- Fix download util
- Fix lambda label misalignment
- Add language pack reference link for SynonymAug
See changelog for more details.
- Data Augmentation library for Text
- Does your NLP model able to prevent adversarial attack?
- How does Data Noising Help to Improve your NLP Model?
- Data Augmentation library for Speech Recognition
- Data Augmentation library for Audio
- Unsupervied Data Augmentation
- A Visual Survey of Data Augmentation in NLP
This library uses data (e.g. capturing from internet), research (e.g. following augmenter idea), model (e.g. using pre-trained model) See data source for more details.
@misc{ma2019nlpaug,
title={NLP Augmentation},
author={Edward Ma},
howpublished={https://github.com/makcedward/nlpaug},
year={2019}
}This package is cited by many books, workshop and academic research papers (70+). Here are some of examples and you may visit here to get the full list.
- S. Vajjala. NLP without a readymade labeled dataset at Toronto Machine Learning Summit, 2021. 2021
- S. Vajjala, B. Majumder, A. Gupta and H. Surana. Practical Natural Language Processing: A Comprehensive Guide to Building Real-World NLP Systems. 2020
- A. Bartoli and A. Fusiello. Computer Vision–ECCV 2020 Workshops. 2020
- L. Werra, L. Tunstall, and T. Wolf Natural Language Processing with Transformers. 2022
- Google: M. Raghu and E. Schmidt. A Survey of Deep Learning for Scientific Discovery. 2020
- Sirius XM: E. Jing, K. Schneck, D. Egan and S. A. Waterman. Identifying Introductions in Podcast Episodes from Automatically Generated Transcripts. 2021
- Salesforce Research: B. Newman, P. K. Choubey and N. Rajani. P-adapters: Robustly Extracting Factual Information from Language Modesl with Diverse Prompts. 2021
- Salesforce Research: L. Xue, M. Gao, Z. Chen, C. Xiong and R. Xu. Robustness Evaluation of Transformer-based Form Field Extractors via Form Attacks. 2021
 sakares saengkaew |
 Binoy Dalal |
 Emrecan Çelik |
