Sentiment Analysis: Opinions Matter, If Only You Knew Which Ones


TOPIC QUESTIONS
Q1) How do you define positive sentiment?
Q2) How does that impact your organizational goals?
Q3) How do you know that what you are measuring matters?
   Event Details
Event Date:
     2010-02-16
Event Time:
     12PM - 1 PM EST
Participants:
     46
Tweets:
     523


sign in with twitter
  Embed:
Moderator
kdpaine.jpg
Katie Delahaye Paine
twitter:
web: www.kdpaine.com
BIO: Katie Delahaye Paine (twitter: KDPaine) is the CEO and founder of KDPaine & Partners LLC and author of, Measuring Public Relationships, the data-driven communicators guide to measuring success. She also writes the first blog and the first newsletters dedicated entirely to measurement and accountability. In the last two decades, she and her firm have listened to millions of conversations, analyzed thousands of articles, and asked hundreds of quest ...more
Filter out Re-tweets
Moderator
10 minutes to today's #socialmedia tweetchat w/ me: topic: sentiment analysis in social media. #sm47 -- is ur Tweetdeck column set up?
#sm47 is about to start. 1st ?? of the day: 1) How do you define positive sentiment?
#SM47 3rd question of the day: 3) How do you know that what you are measuring matters?
#SM47 3rd question of the day: 3) How do you know that what you are measuring matters?
#sm47 would "desirable" be a better term than "Positive" ?
#sm47 and don't forget goals? maybe you want to be invisible??
#sm47 LOL, but that's actually a criteria I hadn't considered
#SM47 and we're off!
#SM47 what about in a crisis, how do you define positive sentiment if you're Toyota + best u can hope for is neutral?
#sm47 yes, definitely !
#sm47 so how do automated sentiment analyzers figure that out?
#sm47 that is perhaps THE question of the day.
#SM47 exactly. it really is more time-defined. If negative goes away in a few days as opposed to a few weeks
#sm47 YES!
#sm47 what about key messages, like "we're fixing the problem" or "We didn't kill THAT many people"
RT : #sm47 Consistency is important as defining sentiment. If "positive" is always applied the same, you can measure & track trends
#sm47 and it doesn't allow for regional differences, e.g. "wicked" is good in this part of the country
#SM47 r u sure? what if it includes both positive and negative?
#sm47 and it doesn't allow for regional differences, e.g. "wicked" is good in this part of the country
#sm47 and how do you measure that?
#sm47 or would you look at "likelihood to purchase or recommend" ?
#sm47 then sentiment should be based on your customers sentiments? or stakeholders? or investors?
#sm47 humans can do both, not sure automated systems can
#sm47 THANK YOU! exactly what I've been finding. Sentiment and SM Metrics are ALL industry and goal specific!
#sm47 thanks for joining. the question is sentiment> How do you define?
#sm47 that's why you need a good human coding system
#sm47 good human coders should have minimum 90% intercoder reliability scores
#sm47 I'd think so,
#sm47 yes, you should be coding post vs comments
#sm47 couldn't have said it better myself
#sm47 or put another way, how do you tie positive sentiment to your bottom line?
Okay folks #SM47, time for the 2nd question of the day: how does your sentiment analysis impact ur organizational goals?
#sm47 yes, you should be coding post vs comments
#sm47 a good option is random sampling
RT : in other words, despite what happened, ppl are likely to purchase or recommend = positive? Think that's appropriate #sm47
#sm47 I think Sysomos does
#Sm47 welcome!
#sm47 hhmmm, not sure I agree with that. Aren't the goals set at the board level for the year?
RT : Q2: some orgs successfully use sentiment of distinct group as canary in coal mine #sm47
#SM47 we call them balanced, as opposed to neutral that have NO sentiment at all
#Sm47 computers have a real hard time with sarcasm, irony and regional dialects
#Sm47 you can't that's why I don't believe its the answer
#sm47 great point. how DO most organizations use it?
#sm47 not specifically, but that's how lots of companies do it
RT : Yet, you must first establish that sentiment analysis and its collection methods can be trusted as "facts" #sm47
#sm47 can you elaborate? do you have examples?
#sm47 is advocacy the goal, or revenue? or membership? doesn't it depend on the type of organization?
#sm47 can you elaborate? do you have examples?
#sm47 but how does that tie to market share or revenue or advocacy?
RT : LOVE IT !90% of media monitoring projects are of such low volume that auto sentiment analysis is a confusing toy. #sm47
#sm47 goals drive metrics - in non-profit everything is positive or at worst neutral, so tool to measure is webanalytics& $$ not sentiment
#sm47 how do you tie sentiment analysis to goals, revenue, sales, market share?
#sm47 yes, can happen within diff deptartment, or diff colleges w/in univ. or diff programs in govt or non-profit
#sm47 classic microsoft moment. Blue screen crash in middle of #sm47
Microsoft uses both human and automated coding :) #sm47
#sm47 I'm back, where was I ?? Question 3: How do you tell whether you are measuring what matters
#sm47 you measure only what matters, and use random sampling when you go over 5000 mentions per month
#sm47 I'm back, where was I ?? Question 3: How do you tell whether you are measuring what matters
#sm47 do all organizations have to worry about scale? humans can efficiently read 5000 mentions a month, how many need more ?
#sm47 I think if they ever did respond, I'd be so shocked I'd have a personal blue screen heart attack.
#sm47 I think if they ever did respond, I'd be so shocked I'd have a personal blue screen heart attack.
#sm47 no, many are just cyberwhiners with no perspective
#sm47 EXACTLY! and how many of you are using correlations to show that?
#sm47 sounds like you've read my book!
RT : When you see improvement toward goal RT : #sm47 Question 3: How do you tell whether you are measuring what matters
#sm47 I'd say velocity + freqeuncy + influence/impact
#sm47 I thinik it is far more important to look for and track trends in complaints than respond to every one
#sm47 which is why you are so desperately needed !
RT : Defining sentiment is as subjective as sentiment itself. #sm47
#sm47 you need to measure whatever medium/forum/network where your customers are. that's all.
#SM47 but actually, I wouldn't measure my own sentiment, I'd measure positive for the competition
#SM47 great question. I'd measure negative
#sm47 isn't that where Board Reader started?
#sm47 proactive has more impact on your bottom line IMHO. reactive is what too many people do
#sm47 isn't that where Board Reader started?
#sm47 proactive has more impact on your bottom line IMHO. reactive is what too many people do
#sm47 depends on the business. Don't think P&G cares about 1 individual sale of a bar of soap
#smbnh #sm47 look forward to see you on Friday!
#sm47 that's why I'm looking at Sysomos and BoardReader, they keep stuff for ever!
#SM47 I want to know what customers perceive that my competition is doing well, so I can improve
#sm47 THANK YOU for the opportunity. Great Convo! great input. if you want to keep this up, just follow me !
#sm47 Yes, and many only get a little piece of Facebook or Twitter, another reason to use humans..
#SM47 its called competitivev intelligence, and I can promise you that Apple does it..
#sm47 I'd kill for a diet coke right now, but actually now buy pepsi because they gave to causes rather than superbowl ads
#SM47 its called competitivev intelligence, and I can promise you that Apple does it..
#SM47 now that's the subject of another chat! or a longer discussion :)
#SM47 did anyone really cancel their reservations because of one guys tweets?
#sm47 that's one of our USPs. :)
#sm47 sorry I have a team of highly efficient fab folks in BerlinNH that I'll put up against computers or anyon in india any day
#sm47 that would be a great test!
#sm47 thanks for the perfect plug for my book "measuring public relationships! " http://bit.ly/6XQjpP
RT : Looking for recap of today's chat on Sentiment Analysis? it's right here http://hashtagsocialmedia.com/event/47. #sm47
#sm47 might I suggest :) http://bit.ly/6XQjpP
#SM47 exactly!
#sm47 why should I send human coding jobs overseas? Don't we need more jobs in the USA?
RT : good on ya! #sm47
thanks everyone, and especially and !! and my Dell XPS 1330 for recovering so quickly :) #sm47