Sentiment analysis -- play the game

[Moving forward with WP5 Tasks 5.2]

Department of Computer Science at North Carolina State University

We have mentioned it would be interesting to provide sentiment analysis feedback to those who would monitor conversations taking place on the forum.

I am interested in finding an appropriate sentiment landscape, I thought the experiment by C. Healy would be worth trying. You enter words, the app scrapes twitter for you and then displays a cloudpoint (points correspond to tweets). Go play!

I am also interested in your feedback about the utility of such a viz. How would you intuitively use such a represntation? Just look at it? Drive the navigation between posts (here tweets) from that viz? Query the posts and get back to the authors’ neighborhood (in the crowd of all authors)? Etc.

Be sentimental

Open your heart.

Imagine you read posts, you are trying to understand what is going on in the Op3nCare crowd, maybe looking how newcomers are doing, maybe looking for places where lively debates take (took) place.

How does the sentiment scale used in the demo serves your search? How would you go from the sentiment map to the data you are looking for?

Come on guys, be generous, tell me all – I need it to fuel WP5. Thanks!

1 Like

Include a customised offer into the fellowship package?

See discussion here?

Not for me

I am no fan of sentiment analysis.

I tried with several words (“edgeryders”, “edgeryders OR opencare”, “stewardship”, but I can’t seem to learn anything of substance.

Hope that was not too disappointing, @melancon

When sentiment analysis hits the fan

@Alberto @MassimoMercuri [Je n’ai pas résisté à faire ce mauvais jeu de mots …]

My guess is you don’t see value in sentiment analysis because up to now you have been able to track almots every and each of the users, and probably every and each post/comment on – this is no surpirse, it’s your job as a community manager!

What if the community grows, what if the volume of excange makes it so that you cannot afford to track each individual or post?

Maybe “sentiment” is not the good way of thinking about how to use this technology. And maybe, it’s true, sentiment coloring is not that useful.

Let’s give it a second chance.

  • What if, on top of the topics that people discuss, I can tag some posts/comments as being “opinions”, “knowledge sharing”, “second”, “contradict”, etc.?

Me, I’m a content freak

I would potentially find it useful only if it can complement the semantic analysis - so on top of us finding what concepts are related and talked about most, we also have a map of feelings around those concepts that puts care priorities in a whole different light. Or the layers you mention (“opinion” “contradict” etc). But if you have an ethnographer analysing the more in-depth conversation, isn’t that covered? @jimmytidey is involved in mapping tweets in online consultation processes, maybe he has some insights for how insightful twitter conversations can be for research purposes?