def read_user_features(about, user_ids, user_id_mapping):
features = Features()
features.item_ids = user_id_mapping
if user_ids:
for uid in user_id_mapping:
features.add_feature(uid, 'user_id:' + uid)
if about:
# Add intercepts
for uid in user_id_mapping:
features.add_feature(uid, 'intercept')
for i, (user_id, about_me) in enumerate(read_user_data()):
clean_about = (strip_tags(about_me)
.replace('\n', ' ')
.lower())
clean_about = _process_post_body(clean_about)
for token in clean_about:
features.add_feature(user_id, 'about:' + token)
features.set_shape()
return features
def read_user_features(about, user_ids, user_id_mapping):
features = Features()
features.item_ids = user_id_mapping
if user_ids:
for uid in user_id_mapping:
features.add_feature(uid, 'user_id:' + uid)
if about:
# Add intercepts
for uid in user_id_mapping:
features.add_feature(uid, 'intercept')
for i, (user_id, about_me) in enumerate(read_user_data()):
clean_about = (strip_tags(about_me).replace('\n', ' ').lower())
clean_about = _process_post_body(clean_about)
for token in clean_about:
features.add_feature(user_id, 'about:' + token)
features.set_shape()
return features
def read_post_features(tags, post_ids, post_text):
features = Features()
for post_id, user_id, post_tags, parent_post_id, body in read_post_data():
# Only get features for questions, not answers.
if parent_post_id is None:
features.add_item(post_id)
if post_ids:
features.add_feature(post_id, 'post_id:' + post_id)
if tags:
for tag in post_tags:
features.add_feature(post_id, 'tag:' + tag)
if post_text:
for token in body:
features.add_feature(post_id, 'body:' + token)
features.set_shape()
return features
def read_post_features(tags, post_ids, post_text):
features = Features()
for post_id, user_id, post_tags, parent_post_id, body in read_post_data():
# Only get features for questions, not answers.
if parent_post_id is None:
features.add_item(post_id)
if post_ids:
features.add_feature(post_id, 'post_id:' + post_id)
if tags:
for tag in post_tags:
features.add_feature(post_id, 'tag:' + tag)
if post_text:
for token in body:
features.add_feature(post_id, 'body:' + token)
features.set_shape()
return features
def read_movie_features(titles=False, genres=False, genome_tag_threshold=1.0, tag_popularity_threshold=30):
features = Features()
with open(os.path.join(DATA_DIR, "movies.dat"), "r") as moviefile:
for line in moviefile:
(iid, title, genre_list) = line.split(SEPARATOR)
genres_list = genre_list.split("|")
features.add_item(iid)
if genres:
for genre in genres_list:
features.add_feature(iid, "genre:" + genre.lower().replace("\n", ""))
if titles:
features.add_feature(iid, "title:" + title.lower())
features.add_title(iid, title)
for iid, tag, relevance in read_genome_tags():
# Do not include any tags for movies not in the 10M dataset
if relevance >= genome_tag_threshold and iid in features.item_ids:
features.add_feature(iid, "genome:" + tag.lower())
# Tags applied by users
## for iid, tag, count in read_tags():
## if count >= tag_popularity_threshold and iid in features.item_ids:
## features.add_feature(iid, 'tag:' + tag)
features.set_shape()
return features
def read_movie_features(titles=False,
genres=False,
genome_tag_threshold=1.0,
tag_popularity_threshold=30):
features = Features()
with open(os.path.join(DATA_DIR, 'movies.dat'), 'r') as moviefile:
for line in moviefile:
(iid, title, genre_list) = line.split(SEPARATOR)
genres_list = genre_list.split('|')
features.add_item(iid)
if genres:
for genre in genres_list:
features.add_feature(
iid, 'genre:' + genre.lower().replace('\n', ''))
if titles:
features.add_feature(iid, 'title:' + title.lower())
features.add_title(iid, title)
for iid, tag, relevance in read_genome_tags():
# Do not include any tags for movies not in the 10M dataset
if relevance >= genome_tag_threshold and iid in features.item_ids:
features.add_feature(iid, 'genome:' + tag.lower())
# Tags applied by users
## for iid, tag, count in read_tags():
## if count >= tag_popularity_threshold and iid in features.item_ids:
## features.add_feature(iid, 'tag:' + tag)
features.set_shape()
return features
