Social media emotions annotation guide (SMEmo): Development and initial validity.

Autor: Paletz SBF; College of Information Studies, University of Maryland, College Park, MD, USA. paletz@umd.edu., Golonka EM; Applied Research Laboratory for Intelligence and Security (ARLIS), University of Maryland, College Park, MD, USA., Pandža NB; Applied Research Laboratory for Intelligence and Security (ARLIS), University of Maryland, College Park, MD, USA.; Program in Second Language Acquisition, University of Maryland, College Park, MD, USA., Stanton G; Department of Criminology, University of Maryland, College Park, MD, USA., Ryan D; Feminist, Gender, and Sexuality Studies, Stanford University, Stanford, CA, USA., Adams N; Applied Research Laboratory for Intelligence and Security (ARLIS), University of Maryland, College Park, MD, USA., Rytting CA; Applied Research Laboratory for Intelligence and Security (ARLIS), University of Maryland, College Park, MD, USA., Murauskaite EE; ICONS Project, University of Maryland, College Park, MD, USA., Buntain C; College of Information Studies, University of Maryland, College Park, MD, USA., Johns MA; Applied Research Laboratory for Intelligence and Security (ARLIS), University of Maryland, College Park, MD, USA., Bradley P; Applied Research Laboratory for Intelligence and Security (ARLIS), University of Maryland, College Park, MD, USA.
Jazyk: angličtina
Zdroj: Behavior research methods [Behav Res Methods] 2024 Aug; Vol. 56 (5), pp. 4435-4485. Date of Electronic Publication: 2023 Sep 11.
DOI: 10.3758/s13428-023-02195-1
Abstrakt: The proper measurement of emotion is vital to understanding the relationship between emotional expression in social media and other factors, such as online information sharing. This work develops a standardized annotation scheme for quantifying emotions in social media using recent emotion theory and research. Human annotators assessed both social media posts and their own reactions to the posts' content on scales of 0 to 100 for each of 20 (Study 1) and 23 (Study 2) emotions. For Study 1, we analyzed English-language posts from Twitter (N = 244) and YouTube (N = 50). Associations between emotion ratings and text-based measures (LIWC, VADER, EmoLex, NRC-EIL, Emotionality) demonstrated convergent and discriminant validity. In Study 2, we tested an expanded version of the scheme in-country, in-language, on Polish (N = 3648) and Lithuanian (N = 1934) multimedia Facebook posts. While the correlations were lower than with English, patterns of convergent and discriminant validity with EmoLex and NRC-EIL still held. Coder reliability was strong across samples, with intraclass correlations of .80 or higher for 10 different emotions in Study 1 and 16 different emotions in Study 2. This research improves the measurement of emotions in social media to include more dimensions, multimedia, and context compared to prior schemes.
(© 2023. The Psychonomic Society, Inc.)
Databáze: MEDLINE
Externí odkaz: