Here is a brief overview of what you need to do: 1. Use nltk to get all the nouns etc. from the topic. 2. You can use LDA/TF-IDF (gensim) for your questions t

Here is a brief overview of what you need to do: 1. Use nltk to get all the nouns etc. from the topic. 2. You can use LDA/TF-IDF (gensim) for your questions to extract the most common topic. 3. Use cosine similarity once you have the corpus built to tag a question to the right category. Look into gensim for more details for the topic modeling.