On: Stanford researchers to open source model they say...

[cantrevealname]

> I'm surprised how well it can parse the sentiment

I tried to fool it, but it took quite a bit of effort to contrive a sentence that it got wrong.

The sentence I finally fooled it with:

--> I enjoy clipping my toenails to paying any more attention to this movie.

which it rated as highly positive.

[andrevoget → cantrevealname]

I could fool the algorithm with: "Only the book cover was good."

[ksrm → cantrevealname]

It does seem to struggle here.

>I'd enjoy clipping my toenails more than paying any more attention to this movie. --> positive

>I'd enjoy clipping my toenails over paying any more attention to this movie. --> negative

If you change "paying" to "giving" it classes that clause as neutral instead of negative, but doesn't change the result.

[eulerphi → cantrevealname]

Your sentence is malformed. It should be "I'd enjoy clipping my toenails to paying any more attention to this movie."

Still, I'm not even sure if using the preposition "to" in that manner is proper english.

[telephonetemp → eulerphi]

It isn't proper English. A proper version of this sentence could be "I'd prefer clipping my toenails to paying any more attention to this movie" but even then the "paying any more attention" part would probably cause a human sentiment analyser to do a double take at it.

[stavros → telephonetemp]

The software above correctly rates this as negative.