-2

Facebook recently reported a controversial experiment in which they found that if people use read words with positive emotional valence in their facebook news feed they will themselves use more positive words in the status updates that they write.

They themselves seem to have used a payed software to do the analysis. Is there a free word list that gives me for all English words a score about their emotional valence?

Christian
  • 788
  • 8
  • 22
  • 3
    Make one up yourself, using any values you please. That's what they did. There are no reliable lists of "emotional valence". Valence is a chemical concept, and gives the impression of concise figures involved (after all, chemical valences are small discrete integers). However, "emotional valence" is only a metaphor, not something one can measure; especially since everybody has different emotions, and uses words in different emotional ways. If, however, you wanna measure them anyway, you hafta do it yourself. – jlawler Jul 22 '14 at 17:17
  • @jlawler : Facebook actually didn't make up a list themselves. That's just not what they did. They didn't use a list they made inhouse.

    We also live in the 21st century. If Facebook would make a list themselves they wouldn't just make up a list by hand but use some machine learning algorithm to create a list.

    I would be surprised if nobody in linguistics or psychology did research on how to set up a decent list that would outperform a list that's naively created by hand.

     – Christian Jul 24 '14 at 15:49
  • It all depends on who coded the training corpus, what they thought they were looking for, and how the corpus was generated. Training up a filter to spot things depends on being able to recognize them accurately in the first place, which means operational definitions. "Operational" for the designers, that is; it has no necessary connection to what anybody else means by the labels they attach to their results. Which is why we know just about everything there is to know about white middle-class American undergraduate students taking psychology classes, and practically nothing about anybody else. – jlawler Jul 24 '14 at 16:33
  • 2
    @Christian How do you know any of these things are true? Does the paper say so? If so, could you include this information in your question? If not, what grounds do you have for assuming it's the case? –  Aug 24 '14 at 05:55
  • @snailboat : I don't actually want to repeat the paper I just want a list of valence of words. Focusing to much on the paper doesn't help with the purpose of the question. As far as to whether or not the Facebook people used an in house solution they say so in the paper in the methods section. – Christian Aug 24 '14 at 08:02

2 Answers2

1

After searching a while I found sentiwordnet which gives for each word a positivity, a negativity and a objectivity rating.

Christian
  • 788
  • 8
  • 22
0

Emotional valence is an extremely complex topic. Emotional state and the expression of such a state is bound by biology, psychology, culture and society. It's methods were not grounded in a consistent self esteem measurements so it's difficult to tell what if anything they proved. They could have just PFUDOR.

The only way to make such a software to quantify emotional state is to have individuals announce their current feelings.

  • That's probably wrong. The measure might not be perfect but that in no way implies that it's worse or equal to random guessing which is what you are effectively argue when you say that the only way to quantify emotional state is to ask individuals. – Christian Jul 24 '14 at 15:24
  • The experiment did pass PNAS peer review. It pretty strange to argue that it's unscientific because you don't understand the methods used. Whether or not the experiment proved what they wanted to prove is an interesting debate but not one in which I'm interested with my question. – Christian Jul 24 '14 at 15:47
  • @Christian link please –  Jul 24 '14 at 15:48
  • Link to what? Are you criticising the study without having read it and want a link from me to read the study? – Christian Jul 24 '14 at 15:50
  • @Christian no I read plenty about the study most of which called it an aberration and unethical the fact that PNAS would publish such pseudo psychology is surprising to me. –  Jul 24 '14 at 15:59
  • In science the usual way to understand a study is not reading something "about the study" but actually reading the paper.

    Being surprised of not understanding why something happens. It's not a good basis for informed criticism even if you don't like the study. As is taking all your information about it from secondary sources instead of reading it directly.

     – Christian Jul 24 '14 at 16:13
  • @Christian unknowing unwilling human subjects is terrible science –  Jul 24 '14 at 16:25
  • 1
    PNAS peer review is a joke. NAS is a profit-making organization devoted to keeping everything behind its paywall. They will publish anything they can sell. – jlawler Aug 23 '14 at 19:31