A framework for comparing stance detection algorithms with 10 different datasets.
On each run, you can select algorithms for comparison and one dataset each time.
The system will display the results of each algorithm and compare the results aswell.
To run the backend please follow the instructions step by step (order is important):
Preparations:
download FNC.csv file from https://drive.google.com/file/d/1pfz4YHjWsglMCctd6RBAFo9kU_y-kMO6/view?usp=sharing and place it in Backend/DB folder
download glove.6b.5d.txt from https://www.kaggle.com/watts2/glove6b50dtxt and place it in Backend/LIU/data folder
download glove.6b.5d.txt from https://www.kaggle.com/watts2/glove6b50dtxt and place it in Backend/SEN folder and Backend/TRANSFORMER/embeddings
download glove.6b.300d.txt from https://www.kaggle.com/thanakomsn/glove6b300dtxt and place it in Backend/SEN folder and Backend/TAN
download glove.6b.zip from https://www.kaggle.com/anindya2906/glove6b and place it in Backend/SEN folder
download snli_1.0.zip from https://nlp.stanford.edu/projects/snli/snli_1.0.zip and place it in the Backend/SEN folder. Extract snli_1.0_dev.txt and place it in Backend/SEN folder
download GoogleNews-vectors-negative300.bin from https://www.kaggle.com/leadbest/googlenewsvectorsnegative300 and place it in Backend/Allada_Nandakumar
Fill tables in DB:
- Go to Backend/DB/DBManager.py and remove comments from commands at the buttom of the page (right after functions declaration).
- Run this file.
- Return the comments (if you won't do it the tables will be duplocated every run).
Run the code:
For backend running, run service.py file
Frontend part:
To run the website you should also clone to frontend repository: https://github.com/irisDreizen/systance-frontend Follow the README there.
Important note:
After your website is up please follow the instructions:
- Go to main page.
- click 'Compare Algorithms' button.
- In datasets - choose semEval2016, In algorithms - choose TAN. Fill 65 for train percent and click on 'Run' botton.
- Wait for the end of the running (that may take a few minutes...).
This part will allow you to run TAN algorithm in 'Check My Stance' part. If you are not willing to use it, you can ignore this note.