diff options
Diffstat (limited to 'data.py')
| -rw-r--r-- | data.py | 5 |
1 files changed, 2 insertions, 3 deletions
@@ -18,13 +18,12 @@ cur.execute('SELECT text FROM statuses WHERE language=\'en\'') statuses = cur.fetchall() +# Clean up statuses for i in range(len(statuses)): # Remove HTML stuff statuses[i] = unescape(sub(r'<[^>]*>', '', statuses[i][0])).split() # Remove URLs and special characters and convert to lowercase - statuses[i] = [sub(r'[^a-z0-9]', '', word.lower()) for word in statuses[i] if word.find('://') == -1] - # Remove empty strings - statuses[i] = ' '.join([word for word in statuses[i] if word != '']) + statuses[i] = ' '.join([sub(r'[^a-z0-9]', '', word.lower()) for word in statuses[i] if word.find('://') == -1]) # Save to output file |