I am trying to build a sentiment analysis proof of concept, and am hitting the wall of getting my data. I found the "German Polarity Clues" dataset, but i can't quite make sense of the data format, can someone help me?

I need to make sense of the "probability" data field. Apparently there are three probabilities included there, but

1) It is not clear what probability we are talking of, is it p(word is contained| positive/negative/neutral sentiment), or is it p(positive/negative/neutral sentiment | word is contained)?

2) There are some negative values as well, what does that mean?

3) In most places there is a "-", but in some there is a 0, what is the distinction?

The dataset can be found here: http://www.ulliwaltinger.de/sentiment/


Feature    (\t)     Lemma     (\t)  Part-of-Speech (\t)  Polarity (\t)  Probability (\t)
hoffnungsloser      hoffnungslos    AD                   negative       -/-0.3412/

Thank you for your help!

  • All the values appear to be between -1 and 1. It's a guess, but I would guess that '-' by itself means no value computed. – Andrew - OpenGeoCode Sep 30 '14 at 12:57
  • I tried writing at the given feedback email address (Ulli) and got "unknown or illegal alias" from their mail server. – Nicolas Raoul Oct 1 '14 at 8:26
  • @NicolasRaoul : it looks like he's moved. If you search on his name, you find ulliwaltinger.de , which states "As of Januar 2012, I am a project manager and research engineer at Siemens Corporate Technology in Munich. Since October 2010 I was working at the Artificial Intelligence Group at Bielefeld University doing my post-doctoral research at the Excellence Cluster…" – Joe Oct 2 '14 at 14:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.