Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

"oh my god i love brazil" is considered negative with high confidence #2

Open
divijan opened this issue Jun 26, 2014 · 6 comments
Open

Comments

@divijan
Copy link

divijan commented Jun 26, 2014

I entered "oh my god i love brazil" into the box on your site, and the result was unexpected:
Result: Negative
Confidence Level: 99.8203

@Marvinsky
Copy link

Where did you put the text "oh my god i love brazil" in order to test it?

Thanks

@divijan
Copy link
Author

divijan commented Jun 26, 2014

At http://sentiment.vivekn.com/.

@iChiragMandot
Copy link

I think, in the training dataset, the occurrences of "oh my god" were used in negative context thereby resulting into such scenario. Also its surprising to see higher accuracy for trigrams as compared to unigram.

@vivekn
Copy link
Owner

vivekn commented Jun 28, 2014

That's correct, "oh my god" has a much greater negative weight than the positive weight of "love" in the training set. That said this model works better on longer sequences of text and doesn't do that well on short phrases.

@divijan
Copy link
Author

divijan commented Jun 28, 2014

Thanks for your replies. To my mind, "oh my god" might only have a negative meaning when used by itself. In conjunction with other phrases it works like a sentiment amplifier.
I mostly deal with short comments in my work, so could you recommend any APIs that work better on short pieces of text?

@hiral2cool
Copy link

if i enter normal text then also said its positive with 100% confidence level
so this is bullshit for NEUTRAL statements

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

5 participants