User-Supplied Sentiments in Tweets

Abstract : Microblogging has become very popular among web users: Twitter broadcasted 200 millions of them every day in 2011. These tweets can be used for social studies, opinion mining, and sentiment analysis. Tweets have been effectively used to analyze opinions and sentiments from specific events such as TV political debates or conference presentations. To analyze tweets, researchers either use Natural Language Process- ing or human-analyses, e. g. with Amazon Mechanical Turks. In this paper we describe a third method that can replace or complement the first two: convincing the tweet authors to explicitly express their opinion using a simple unambiguous syntax--a tag--while they tweet. We report on our experience with the PolemicTweet system where authors tag the tweets they send during conferences and TV shows. We explain how we implemented PolemicTweet and discuss briefly the pro and cons. We believe that author-tagging technologies can effectively complement automated analysis, in particular for analyzing tweets that are by essence ambiguous, and could be ironic, sarcastic or cryptic. We describe the incentive we used to convince the authors and the result we obtained.
Document type :
Conference papers
Complete list of metadatas

Cited literature [18 references]  Display  Hide  Download

https://hal.inria.fr/hal-00734407
Contributor : Samuel Huron <>
Submitted on : Friday, September 21, 2012 - 4:54:33 PM
Last modification on : Thursday, February 9, 2017 - 3:47:42 PM
Long-term archiving on : Friday, December 16, 2016 - 3:12:50 PM

File

user-supplied_sentiments-in-tw...
Files produced by the author(s)

Identifiers

  • HAL Id : hal-00734407, version 1

Collections

Citation

Samuel Huron, Jean-Daniel Fekete. User-Supplied Sentiments in Tweets. IEEE Conference on Information Visualization (INFOVIS): 2nd Workshop on Interactive Visual Text Analysis, IEEE, Oct 2012, Seattle, Washington, USA, United States. ⟨hal-00734407⟩

Share

Metrics

Record views

400

Files downloads

285