rosscg / newsworthy Goto Github PK

This app currently takes a Twitter user's news feed and evaluates the extent to which the articles they share agree or disagree with articles published by six major news outlets.

It uses a classification algorithm developed by UKP Lab (https://github.com/hanselowski/athene_system)

==================================

Install python3. Clone this repository. Create virtual environment.

python3 -m venv myvenv
source venv/bin/activate

Install requirements.

git clone "https://github.com/hanselowski/athene_system.git"
python3 pip install -r requirements.txt

Edit feed_compare/tokens_SKELETON.py to include relevant Twitter tokens (Consumer Key, Secret, and Access Token), and rename to tokens.py.

Unzip the features file to the directory:

athene_system/data/fnc-1/features

Unzip the model file to the directory:

athene_system/data/fnc-1/mlp_models

Parts of the Natural Language Toolkit (NLTK) might need to be installed manually:

python3 -c "import nltk; nltk.download('stopwords'); nltk.download('punkt'); nltk.download('wordnet')"

Optional: Copy Word2Vec GoogleNews-vectors-negative300.bin.gz to folder athene_system/data/embeddings/google_news/.

Optional: Download Paraphrase Database: Lexical XL Paraphrases 1.0 and extract it to the athene_system/data/ppdb folder.

==================================

python app.py

Browse to http://localhost:5001/