def get_language_file_name(app_id=None):
if app_id is None:
app_id = get_artifact_app_id()
file_extension = '.json'
language_file_name = get_language_folder_name() + str(
app_id) + file_extension
return language_file_name
def main():
# Pre-processing (only required once)
apply_language_detector()
# Processing: Artifact
app_id = get_artifact_app_id()
english_review_dict = filter_reviews(app_id)
return
def download_app_details(app_id=None):
# For app_id, download app details (including store description)
if app_id is None:
app_id = get_artifact_app_id()
app_id = str(app_id)
(app_details, _, _) = steampi.api.load_app_details(app_id)
return app_details
def filter_reviews(app_id=None):
if app_id is None:
app_id = get_artifact_app_id()
english_review_dict = load_english_reviews(app_id)
english_review_dict = filter_out_short_reviews(english_review_dict,
length_threshold=150)
print('#reviews = {}'.format(len(english_review_dict)))
detected_languages = detect_review_language(app_id)
english_review_dict = filter_out_reviews_not_detected_as_english(
english_review_dict, detected_languages)
print('#reviews = {}'.format(len(english_review_dict)))
return english_review_dict
def main():
app_ids = [get_artifact_app_id()]
data_request = dict()
data_request['request'] = 'top100in2weeks'
data = steamspypi.download(data_request)
app_ids += list(data.keys())
for app_id in app_ids:
english_review_dict = filter_reviews(app_id)
concatenated_reviews = concatenate_reviews_as_a_large_text(
english_review_dict)
with open(get_output_file_name(app_id), 'w', encoding='utf8') as f:
print(concatenated_reviews, file=f)
return
except FileNotFoundError:
print('Model not found. Setting initial epoch to 0.')
initial_epoch = 0
model.fit(x,
y,
batch_size=128,
epochs=num_epochs,
initial_epoch=initial_epoch,
callbacks=[print_callback, save_callback])
return model
if __name__ == '__main__':
app_id = get_artifact_app_id()
text = read_input(get_output_file_name(app_id))
maxlen = 20
params = get_params(maxlen)
text = trim_text(text, params['chars'])
model = train_model(text,
maxlen,
num_epochs=20,
full_model_filename=None,
initial_epoch=0)
start_index = random.randint(0, len(text) - maxlen - 1)
for diversity in [0.2, 0.5, 1.0, 1.2]: