def load_poems_model(file_name, w2v_model, vectorize=True):
pmodel = read_data_model(file_name)
print("loading model...")
if vectorize:
print("vectorizing model...")
pmodel['matrices'] = [sem.bag_to_matrix(bag, w2v_model) for bag in pmodel['bags']]
pmodel['a_matrices'] = [sem.bag_to_matrix(bag, w2v_model) for bag in pmodel['associations']]
print("poems model '%s' loaded" % file_name)
return pmodel
def load_poems_model(file_name, w2v_model, vectorize=True):
pmodel = read_data_model(file_name)
print("loading model...")
if vectorize:
print("vectorizing model...")
pmodel['matrices'] = [sem.bag_to_matrix(bag, w2v_model) for bag in pmodel['bags']]
pmodel['a_matrices'] = [sem.bag_to_matrix(bag, w2v_model) for bag in pmodel['associations']]
print("poems model '%s' loaded" % file_name)
return pmodel
def similar_poems_idx(query: str, poem_model, w2v_model, topn=5, use_associations=False) -> list:
query_bag = sem.canonize_words(query.split())
if use_associations:
query_bag += sem.semantic_association(query_bag, w2v_model, topn=5)
query_mx = sem.bag_to_matrix(query_bag, w2v_model)
similars = [(i, sem.semantic_similarity_fast(query_mx, np.vstack((mx, poem_model['a_matrices'][i]))))
for i, mx in enumerate(poem_model['matrices']) if len(mx) > 0]
else:
query_mx = sem.bag_to_matrix(query_bag, w2v_model)
similars = [(i, sem.semantic_similarity_fast(query_mx, mx))
for i, mx in enumerate(poem_model['matrices'])]
similars.sort(key=lambda x: x[1], reverse=True)
return similars[:topn]
def similar_poems_idx(query: str, poem_model, w2v_model, topn=5, use_associations=False) -> list: # [(poem_idx, sim)]
query_bag = sem.canonize_words(query.split())
if use_associations:
query_bag += sem.semantic_association(query_bag, w2v_model, topn=5)
query_mx = sem.bag_to_matrix(query_bag, w2v_model)
if len(query_mx) == 0:
return []
similars = [(i, sem.semantic_similarity_fast(query_mx, np.vstack((mx, poem_model['a_matrices'][i]))))
for i, mx in enumerate(poem_model['matrices']) if len(mx) > 0]
else:
query_mx = sem.bag_to_matrix(query_bag, w2v_model)
if len(query_mx) == 0:
return []
similars = [(i, sem.semantic_similarity_fast_log(query_mx, mx))
for i, mx in enumerate(poem_model['matrices'])]
# similars.sort(key=lambda x: x[1], reverse=True)
return heapq.nlargest(topn, similars, key=lambda x: x[1])
