[Corpora-List] Meta Question regarding sentiment/opinion/reputation was Re: Campagne d=?windows-1252?Q?=92=E9tiquetage_?=en sentiments de tweets et blogs

[Trevor Jenkins]

On 19 Nov 2012, at 15:00, Patrick Juola <juola at mathcs.duq.edu> wrote:

> On Mon, Nov 19, 2012 at 8:35 AM, Trevor Jenkins <trevor.jenkins at suneidesis.com> wrote:
> On 19 Nov 2012, at 12:44, <coralie.reutenauer at orange.com> wrote:
> 
>> Here is a web site for sentiment tagging of tweets and blogs, where you can analyze brand e-reputation :
> 
> I've been pondering an issue with sentiment analysis and opinion mining that the above brings to a head. Are the three things: sentiment, opinion and reputation identical?
> 
> No, they're not.   Consider the following exchange:
> - Have I eaten my last Twinkie?
> - Unfortunately, yes.   

But that's a culturally determined example. Twinkies, thankfully, never made it across the Atlantic.

> The second speaker is expressing an opinion (that no one else will pick up the Twinkie brand and start manufacturing them again, which is value neutral), a sentiment (negative) that this is a bad thing, and a reputation (positive) for Twinkie, because its disappearance is bad.

Alternatively the second speaker is expression a sentiment that the first speaker has wolf down the last of the Twinkies in the house and that they are resentful of the action. Without a (much) larger context we can't read into that anecdotal example any real sentiment, opinion or reputation.
>  
> Are we devaluing each or any of those descriptors by regarding them as one and same lumping them together like this?
> 
> No, because they're very similar.   "Big cat" is a meaningful term even if lions, tigers, and panthers are all distinct species, because they share more similarities than differences.

Not it you're an Apple user. Hopefully I can out run those. ;-)

Regards, Trevor.

<>< Re: deemed!

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/corpora/attachments/20121119/637218c7/attachment.htm>
-------------- next part --------------
_______________________________________________
UNSUBSCRIBE from this page: http://mailman.uib.no/options/corpora
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora