vanzytay / Quaterniontransformers
Repository for ACL 2019 paper
Stars: ✭ 56
Programming Languages
python
139335 projects - #7 most used programming language
Projects that are alternatives of or similar to Quaterniontransformers
Corenlp
Stanford CoreNLP: A Java suite of core NLP tools.
Stars: ✭ 8,248 (+14628.57%)
Mutual labels: natural-language-processing
Thot
Thot toolkit for statistical machine translation
Stars: ✭ 53 (-5.36%)
Mutual labels: natural-language-processing
Vietnamese Electra
Electra pre-trained model using Vietnamese corpus
Stars: ✭ 55 (-1.79%)
Mutual labels: natural-language-processing
Lingua Franca
Mycroft's multilingual text parsing and formatting library
Stars: ✭ 51 (-8.93%)
Mutual labels: natural-language-processing
Fasttext multilingual
Multilingual word vectors in 78 languages
Stars: ✭ 1,067 (+1805.36%)
Mutual labels: natural-language-processing
Market Reporter
Automatic Generation of Brief Summaries of Time-Series Data
Stars: ✭ 54 (-3.57%)
Mutual labels: natural-language-processing
Convai Baseline
ConvAI baseline solution
Stars: ✭ 49 (-12.5%)
Mutual labels: natural-language-processing
Research papers
Record some papers I have read and paper notes I have taken, also including some awesome papers reading lists and academic blog posts.
Stars: ✭ 55 (-1.79%)
Mutual labels: natural-language-processing
Notes
The notes for Math, Machine Learning, Deep Learning and Research papers.
Stars: ✭ 53 (-5.36%)
Mutual labels: natural-language-processing
Emotion Detector
A python code to detect emotions from text
Stars: ✭ 54 (-3.57%)
Mutual labels: natural-language-processing
Iob2corpus
Japanese IOB2 tagged corpus for Named Entity Recognition.
Stars: ✭ 51 (-8.93%)
Mutual labels: natural-language-processing
Python Tutorial Notebooks
Python tutorials as Jupyter Notebooks for NLP, ML, AI
Stars: ✭ 52 (-7.14%)
Mutual labels: natural-language-processing
Jieba Php
"結巴"中文分詞:做最好的 PHP 中文分詞、中文斷詞組件。 / "Jieba" (Chinese for "to stutter") Chinese text segmentation: built to be the best PHP Chinese word segmentation module.
Stars: ✭ 1,073 (+1816.07%)
Mutual labels: natural-language-processing
Spark Nkp
Natural Korean Processor for Apache Spark
Stars: ✭ 50 (-10.71%)
Mutual labels: natural-language-processing
Coarij
Corpus of Annual Reports in Japan
Stars: ✭ 55 (-1.79%)
Mutual labels: natural-language-processing
Pattern
Web mining module for Python, with tools for scraping, natural language processing, machine learning, network analysis and visualization.
Stars: ✭ 8,112 (+14385.71%)
Mutual labels: natural-language-processing
Nltk Book Resource
Notes and solutions to complement the official NLTK book
Stars: ✭ 54 (-3.57%)
Mutual labels: natural-language-processing
Hmtl
🌊HMTL: Hierarchical Multi-Task Learning - A State-of-the-Art neural network model for several NLP tasks based on PyTorch and AllenNLP
Stars: ✭ 1,084 (+1835.71%)
Mutual labels: natural-language-processing
Demos
Some JavaScript works published as demos, mostly ML or DS
Stars: ✭ 55 (-1.79%)
Mutual labels: natural-language-processing
Scdv
Text classification with Sparse Composite Document Vectors.
Stars: ✭ 54 (-3.57%)
Mutual labels: natural-language-processing
QuaternionTransformers
This is our Tensor2Tensor implementation of Quaternion Transformers. This paper will be presented in the upcoming ACL 2019 in Florence.
Dependencies
- Tensorflow 1.12.0
- Tensor2Tensor 1.12.0
- Python 2.7
Usage
- The usage of this repository follows the original Tensor2Tensor repository (e.g., t2t-datagen, t2t-trainer followed by t2t-decoder). It helps to gain familiarity on T2T before attempting to run our code.
- Setting
--t2t_usr_dir=./QuaternionTransformers
will allow T2T to register Quaternion Transformers. To verify, usingt2t-trainer --registry_help
to verify that you are able to load Quaternion transformers. - You should be able to load
MODEL=quaternion_transformer
and use base or big setting as per normal. - Be sure to set
--hparams="self_attention_type="quaternion_dot_product""
to activate Quaternion Attention. - By default, Quaternion FFNs are activated for positional FFN layers. To revert and not use Quaternion FFNs on the position-wise FFN, set
--hparams="ffn_layer="raw_dense_relu_dense"
.
Citation
If you find our work useful, please consider citing our paper:
@article{tay2019lightweight,
title={Lightweight and Efficient Neural Natural Language Processing with Quaternion Networks},
author={Tay, Yi and Zhang, Aston and Tuan, Luu Anh and Rao, Jinfeng and Zhang, Shuai and Wang, Shuohang and Fu, Jie and Hui, Siu Cheung},
journal={arXiv preprint arXiv:1906.04393},
year={2019}
}
Note that the project description data, including the texts, logos, images, and/or trademarks,
for each open source project belongs to its rightful owner.
If you wish to add or remove any projects, please contact us at [email protected].