traindf['title'] =traindf['title'].apply(lambda x: ' '.join([re.sub('[^a-zA-Z]' , "",word) for word in x.split() ]))