Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
Browse files

Final change for the report.

  • Loading branch information...
commit bbcefdeda816e019d3c0d940e43ba437b329be73 1 parent 9b1b63c
@rxin authored
Showing with 1 addition and 1 deletion.
  1. +1 −1  doc/report.tex
View
2  doc/report.tex
@@ -127,7 +127,7 @@ \subsection{Parameter Tuning}
\end{figure}
-\textbf{Stemming}: We observed that stemming the words had a negative impact on the F1 score for both multinomial or Bernoulli models, while using more computational resources. We speculate that the choices of variations of words indeed carry connotations on sentiments. This observation is consistent with findings from previous works \cite{stanford-tutorial, sentiment-twitter}.
+\textbf{Stemming}: We observed that stemming the words had a negative impact on the F1 score for both multinomial or Bernoulli models. We speculate that the choices of variations of words indeed carry connotations on sentiments. This observation is consistent with findings from previous works \cite{stanford-tutorial, sentiment-twitter}.
\textbf{multinomial or Bernoulli}: We observed that multinomial models perform better than Bernoulli models. This is expected as movie reviews are long documents. Based on our understanding, multinomial models work better for longer documents, while Bernoulli models work better for very short documents.
Please sign in to comment.
Something went wrong with that request. Please try again.