Open BenjaminStorer opened 5 years ago
I have some thoughts on this, might be something for me to tackle as it starts going into the Firebase backend away from Flutter
Conducting some initial investigation into using a language vector space and vector word embeddings to be the basis of the interest matching algorithm.
Initial experiments using a 40,000 size word embedding dictionary have given somewhat positive results (but also some unexpected behaviour).
Moving this out of scope to alpha release - will be reassigned after alpha release
Implementing a system which prioritises more related interests when filtering through users to match.