In this repo, I will implement some NLP models for the nlp beginner
learner. In each project folder contains a notebook floder to show the
output in detail. I think this would help the beginner to understand
what happens in the model.
I will list what resource used for each model implementation. All
project based on Python3.6 and Keras2.1.6 with TensorFlow1.8 backend.
Deep Models for NLP beginners
You can find detail instruction in each project. Here I will list what you can learn in each project.
In this project, I use three embedding levels,
word/character/subword, to represent the text. And test them with two
model, CNN and LSTM.
According to the result, subword-level embedding is useful for the
dataset with many unknown words. The CNN not only achieve the better
performance, but also take less training time. So if you want to
implement a simple and powerful sentiment classification model, I highly
recommend to use the CNN model.
I use conda to construct the environment, and I highly
recommend you do it too. After clone this project, you can run the
following command to construct the whole environment. Make sure you
already install the conda tool.
conda env create -f py36.yml
After install the whole environment, you can use following command to switch to the py36 environment.
source activate py36
Bibtex
Please use the following bibtex, when you refer my implementations in your papers.
@misc{liang2018kerasnlpmodel,
title = {Keras Implementations for NLP Models},
author = {Xu, Liang},
url = {https://github.com/BrambleXu/nlp-beginner-guide-keras},
year = {2018}
}