A robot for detecting BS statements
Currently this works as a Twitter bot. Mention @BSORNOT1 in order to get an assessment of whether the statement is bullshit or not.
To do:
- collect data for testing metrics
- develop the other metrics
- start making use of Google NLP API
- create startup script (so that when machine restarts service is not interrupted)
- add a REST API to allow other services to ask questions to the robot.
- add a streamer to Telegram
- add a way to receive feedback from authorised users
- add a way to measure confidence level of assessment
- add a way to improve the model with increased feedback
In order to test the system use:
python -m unittest discover -v
To run the Twitter streamer use:
python -u bsornot/bot.py
Note that the source code does not include the file secrets.py, or other account secrets, for obvious reasons. That means you will need to mock it in order to run it locally. There's no reason for you to want to test the streamer though (well, not yet anyway... we are nowhere near close to have to worry about scalability). The NLP APIs are a different story, so in order to test the system locally you will need to have your own mock project on Google Cloud.