video_recommender.py
670 Bytes
from gensim.models import Word2Vec
import numpy as np
def recommend_videos(tags, tag_model_path, video_model_path, top_k):
tag_vectors = Word2Vec.load(tag_model_path).wv
video_vectors = Word2Vec().wv.load(video_model_path)
error_tags = []
video_vector = np.zeros(100)
for (tag, weight) in tags:
if tag in tag_vectors.vocab:
video_vector = video_vector + (tag_vectors[tag] * float(weight))
else:
# Pass if tag is unknown
if tag not in error_tags:
error_tags.append(tag)
similar_ids = [x[0] for x in video_vectors.similar_by_vector(video_vector, top_k)]
return similar_ids