Beyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Fake News Checker Chatbot: MisterSpocker
1. Fake News Checker Chatbot:
MisterSpocker
Global AI Hackathon 2017
Team Members:
Aki Kutvonen
Supachai C
Joey Fuentes
2. How It Works
• Facebook News Posts
• Data Set
• BuzzFeedNews data set
• Processes
• Data mining from news in Facebook
• Collect inputs
• Facebook posts comments (likes, shares, actual text)
• News article text
• Comparison with other reliable news sources (CNN, BBC, etc.)
• NLP: Bigrams & TD-IFD, Sentiment Analysis, etc.
• Random forest, Linear regression, Neural Networks, Stacking, Ensembling, etc.
• Comparison with other reliable news providers
• Chatbot interface
• Output:
• Falseness and factualness rating
3. Data Set
• BuzzFeedNews Dataset CSV file:
• https://github.com/BuzzFeedNews/2016-10-facebook-
fact-check/blob/master/data/facebook-fact-check.csv
4. Processes
• Data mining from news in Facebook
• Collect inputs
• Facebook posts comments (likes, shares, actual text)
• News article text
• Comparison with other reliable news sources (CNN, BBC, etc.)
• NLP: Bigrams & TD-IFD, Sentiment Analysis, etc.
• Random forest, Linear regression, Neural Networks, Stacking, Ensembling, etc.
• Comparison with other reliable news providers
• Chatbot interface
• NLP:
• sentiment analysis of news reader comments and reactions.
• using Google Search API on News - search using news headline if it exist in a reliable news provider: CNN, BBC.
• Chatbot creation - copy paste Facebook news URL to check news.
5. Output
• Based on the Facebook URL input, Facebook chatbot
replies with degrees of news falseness and factualness
7. MisterSpocker chatbot sample:
“Hello, please enter news URL”
You: enter URL
“This news is probably false (0.80).
Also it is non-factual (0.30).”
“This news might be related:
www.cnn…”