[Corpora-List] EmoText - Software for opinion mining and lexical affect sensing

[Robert Zimbardo]

Given the performance that the system has been shown to exhibit in
just these few examples discussed on the list, I think it has very few
if any "undisputable advantages" (to use your own strange marketing
speak). What it boils down to is:

1) you're trying to sell a 'product' that produces results that
produce very counterintuitive results
2) you're explaining part of the 'product's' bad performance by
pointing out its inability (!) to do something any such software
should be able to do, namely decompose completely compositional words
it doesn't know (such as "unimpressed")
3) finally, you are bold enough to market something like this to
industry customers without mentioning its level of development while,
on the other hand, on this list, you write "BTW I also want to learn
something. :)", which makes it clear to us, but not to potential
customers who might be tempted to pay a lot of money for this
undisputably advantageous software, that this is beta at best

With all respect that may be due to your Ph.D. thesis, which I haven't
read, but this 'website/product' does a disservice to the fields of
statistical NLP, sentiment analysis, etc.
RZ

_______________________________________________
UNSUBSCRIBE from this page: http://mailman.uib.no/options/corpora
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora