readme

Vectorization options:
SIMPLER
- TF-IDF 
- count vectorizer
ADVANCED
- doc2vec
- word2vec
- BERT embeddings
- GloVe

model options:
- semantec search
- BERT (?)
- knn w/ euclidian distance
- Cosine similarity, euclidian distance (?)


NEAL:
TFIDF, BERT

DAN:
GLOVE embeddings, word2vec, fasttext

TROY:
ELMO, ULMfit, doc2vec

We need to create a vector database to store the text data once it is vectorized. (unsure if this is some type of special file or something).

HYBRID APPRAOCH:
    This is how we layer on the color, type, and other attributes. Things like color would be used as an initial screen to filter the cards, if there is a secondary attribute or something (like type? cost?) 
then we could vectorize that if need be (or just change that to a number) and add it to the search. We would weight the relevent pieces to create an overall similarity score.

TEST CARDS:
Thorbran, Thane of Red Fell
Rocco, Street Chef
Sai, Master Thopterist