Put video fies in /videos
Add video name, file name, and description to the videos dictionary in nlp.py
Using CMU's pocketphix for keyword listening (see listening.py): https://github.com/cmusphinx/pocketsphinx-python
sudo pip3 install pocketsphinx
And using the speech recognition library for the google speech api after keyword: https://github.com/Uberi/speech_recognition#readme
sudo pip3 install SpeechRecognition
Using spaCy for speech phrase vs. video description similarity: https://github.com/explosion/spaCy
sudo pip3 install -U spacy
For the raspberry pi there is not enought memory to make spacy. Follow these steps to update swap file size, boot options, and memory split:
sudo nano /etc/dphys-swapfile
change CONF_SWAPSIZE=100
to CONF_SWAPSIZE=1024
sudo /etc/init.d/dphys-swapfile stop
and sudo /etc/init.d/dphys-swapfile start
sudo raspi-config
choose the console option via Boot Options => Desktop / CLI => Console Autologin
sudo raspi-config
choose advanced Options => Memory Split and set to 16
DON'T FORGET TO UNDO THESE STEPS AFTER INSTALL
large swap sizes corrupt memory cards :) and we need the GPU for the videos later.
sudo python3 -m spacy download en