Q$2$: Improving LDA through better TFIDF model $[6$ points$]$

Now, looking at your answers from Q$1,$ you will see that the topics printed by print$\_$top$\_$words dont make much sense. Lets look at a few ways

we can improve this. First challenge is that the amount of tokens on the full data is large. Also we have a lot of stop words,

Task:

Let us limit the TfidfVectorizer to about $5000$ tokens $($max$\_$features$)$ and set the Tfidfvectorizer to remove english stopwords.

Save the new vectorizer as vectorizer$\_$tfidf$\_$lim

# $2$ Points

# YOUR CODE HERE

raise NotImplementedError$()$

vectorizer$\_$tfidf$\_$limit$.$ fit$($documents$)$

tfidf$\_$feature names$\_$limit$\_$ vectorizer$\_$tfidf$\_$limit$.$ get$\_$feature$\_$names$\_$out$()$

Ida$\_$tfidf$\_$limit $=$ fit$\_$LDA$($X$\_$dtm$\_$tfidf$\_$limit$,n_{-}$components