def get_job_embs(
title_id,
emb_dim=300,
model_path="/data/rali7/Tmp/solimanz/data/wikipedia/wiki.en.bin"):
"""
Returns a matrix of job title embeddings
"""
print("Loading fastText model...")
model = FastText(model_path)
print("Model successfull loaded :-D")
embeddings = np.zeros((len(title_id), emb_dim), dtype=np.float32)
for title in title_id.keys():
vec = model.get_sentence_vector(title)
embeddings[title_id[title], :] = vec
return embeddings
train_targets)
np.save(os.path.join(ds_path, "bow", "test_targets.npy"),
test_targets)
print("All done! :-D")
if args.representation == 'title_emb' or args.representation == 'all':
print("Loading pre trained fastText model...")
model = FastText(
"/data/rali7/Tmp/solimanz/data/wikipedia/wiki.en.bin")
X_train = np.zeros((len(train), 300), dtype=np.float32)
X_test = np.zeros((len(test), 300), dtype=np.float32)
for i, job_hist in enumerate(train):
vec = model.get_sentence_vector(job_hist)
X_train[i, :] = vec
for i, job_hist in enumerate(test):
vec = model.get_sentence_vector(job_hist)
X_test[i, :] = vec
create_path(os.path.join(ds_path, 'emb'))
print("Saving train and test embeddings...")
np.save(os.path.join(ds_path, "emb", "train.npy"), X_train)
np.save(os.path.join(ds_path, "emb", "test.npy"), X_test)
np.save(os.path.join(ds_path, "emb", "train_targets.npy"),
train_targets)
np.save(os.path.join(ds_path, "emb", "test_targets.npy"),
test_targets)
