Inspired by the paper from Meta Research lab, this project is a try to build a NLP model that embeded End to End Network and LSTM techniques.
Through training phase, the model tried to catch the relation between the words that it saw in training dataset.
And when making the prediction, it will calculate the probobility of 'Yes' and 'No' among question words.
If the probobility of 'Yes' is greater than 'No', the answer will be positive.
Link of live demo page can be found at:
https://rexxwei.github.io/portfolio/
To make this project work, below libraries or features must included in your Python environment.
- Flask
- SciKit-Learn
- Pandas
- Beautifulsoup4
- Jinja2
- Joblib
- matplotlib
- Numpy
- Scipy
- Seaborn
The dependence can be installed by execute below command
pip install -r requirements.txt
In the project directory, run the Python file 'app.py'.
python app.py
Then open a browser and try the address like:
localhost:8080
Due to the small vocabulary size in the dataset, the words never shown up in the vocabulary won't be recoganized.
Hope a bigger size vocabulary dataset will be available soon.