Crowd sentiment detection during disasters and crises

From Brede Wiki

Jump to: navigation, search
Conference paper (help)
Crowd sentiment detection during disasters and crises
Authors: Ahmed Nagy, Jeannie Stamberger
Citation: Proceedings of the 9th International ISCRAM Conference  : 2012 April
Editors: L. Rothkrantz, J. Ristvej, Z. Franco
Publisher: Define publisher
Meeting: 9th International ISCRAM Conference
Database(s): Google Scholar cites
DOI: Define doi.
Link(s): https://www.cmu.edu/silicon-valley/dmi/files/crowd_sentiment_detection.pdf
Search
Web: DuckDuckGo Bing Google Yahoo!Google PDF
Article: Google Scholar PubMed
Restricted: DTU Digital Library
Services
Format: BibTeX

Crowd sentiment detection during disasters and crises describes a text sentiment analysis of Twitter messages

[edit] Data

3698 tweets in regards the the "San Bruno event" in the first 24 hours. They used TwapperKeeper based on sanbrunofire keyword.

[edit] Method

Crowdflower [1] was use to manually classify tweets.

For text sentiment analysis SentiWordNet, emoticons and the AFINN word list was used as well as a "Bayesian Network".

[edit] Criticism

  1. The AFINN word list is misspelt as AFNN. In the code even AKNN.
  2. The AFINN word list is described as having values fromo -3 to +3. Actually the values are from -5 to +5.
  3. The conclusion is unclear. It states " Using Bayesian Networks with SentiWordNet yielded the best precision and recall" However the F-measure in Table 3 seems to show that the combination of AFINN, emoticons, SentiWordNet and the Bayesian network is the best.
Personal tools