Quotes are proposed dependent on the sentiment scores of the `nltk` library. It seems like it needs some improvement to return quotes which are more similar to the sentence the user typed in.