An Implementation of ULMFit
Twitter US Sentiment Data can be downloaded via the site.
'twitter_data_exploration.ipynb' gives you a nice overview of the data exploration and an implementation of ULMFit to obtain an accuracy of 62%
A detailed overview is provided by Jeremy Howard in the fastai forum
On Mac:
- Open Terminal
- Navigate to desired folder: cd "~\path\folder"
- Clone fastai repository from GitHub: git clone https://github.com/fastai/fastai.git
- Navigate to cloned repository: cd fastai
- Create environment: conda env create -f environment-cpu.yml
- Activate environment: conda activate fastai-cpu
- Start Jupyter Notebook: jupyter notebook
On Windows: The Prompt console must be runned as an administrator. Commands must be run in the Anaconda Prompt console.