-1
$\begingroup$

I'm working with datumbox ml framework for sentiment analysis.Here the git link. I found a sentiment Ml code in code examples of frame work here is the code example link. I want to use same code for twitter-sentiment But in the code example,only two training datasets are used pos,neg But for Twitter sentiment analysis,I want to include Neutral sentiment also.I do have my own training datasets of pos & neg tweets but I don't any Neutral datasets.

My doubt is :

1) Do I need to feed neutral training dataset for giving the neutral sentiment? or

2) Do I need to predict neutral sentiment based on positive & negative training sets.if I need go for predicting then I need an idea on how to do that?

$\endgroup$
1

1 Answer 1

0
$\begingroup$
  1. The machine cannot return neutral sentiment if you do not feed it. If you just have positive or negative, the machine will not return neutral.
  2. Alternatively, maybe you can try other approaches. You can check the confidence with which it predicts positive or negative and if it's low, maybe you can assign it to neutral.
$\endgroup$
5
  • $\begingroup$ can share any code example @Hima Varsha $\endgroup$ Commented Jul 14, 2016 at 6:37
  • $\begingroup$ umm, I have not come across codes for second option I suggested. There are research papers though which do sentiment analysis on Twitter with 3 classes - positive, negative and neutral. I am sure there will be datasets as well. Would you want that? $\endgroup$ Commented Jul 14, 2016 at 7:12
  • $\begingroup$ yeah ! it will be helpful @Hima Varsha $\endgroup$ Commented Jul 14, 2016 at 8:25
  • $\begingroup$ www-nlp.stanford.edu/courses/cs224n/2009/fp/3.pdf this actually contains datasets, codes and loads of references. Also, there is crowdsourcing-class.org/assignments/downloads/pak-paroubek.pdf which contains all 3 classes. $\endgroup$ Commented Jul 14, 2016 at 12:41
  • $\begingroup$ hello can you help me to get answer for this question $\endgroup$ Commented Jul 22, 2016 at 9:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.