laughbot

Predicting Humor in Spoken NLP

Training

• If no .test/val/train files, run python createbaselinetrain.py
• To run regression model: python baseline.py

Realtime

Run python laughbot_realtime.py

Using Text-to-Speech transcription:

Setup Google Cloud API

• Follow https://cloud.google.com/speech-to-text/docs/quickstart-protocol
• Create a service account, save locally only as [NAME].json
• Run export GOOGLE_APPLICATION_CREDENTIALS=./[NAME].json
  • you will need to do this everytime you open a new shell session
• Install Google Cloud SDK: https://cloud.google.com/sdk/downloads
  • when it asks to create bash profile and leave blank or provide path, use the path it suggests but change .bash_profile to .profile (https://stackoverflow.com/questions/31084458/installed-google-cloud-sdk-but-cant-access-gcloud)
• Run gcloud auth application-default login
  • Login to google cloud console
• Run gcloud auth activate-service-account --key-file=./service_account_key.json
  • enable service account

Details

• Commands will reference project laughbot by default
• Compute Engine commands will use region us-west1 by default
• Compute Engine commands will use zone us-west1-b by default

Imports

• pip install SpeechRecognition (https://pypi.python.org/pypi/SpeechRecognition/)
• pip install google-api-python-client
• brew install portaudio && brew link portaudio
• pip install pyaudio