Twitter, time and emotions.

Area of Science:

• Computational social science
• Affective science
• Natural language processing

Background:

• Emotions expressed in tweets exhibit circadian (24-hour) and circaseptan (7-day) patterns.
• Previous analyses relied solely on text-based sentiment analysis, potentially limiting findings.
• Self-referential content holds greater personal meaning, yet its temporal emotional expression is understudied.

Purpose of the Study:

• To compare temporal emotion patterns in self-referencing versus other tweets.
• To contrast sentiment analysis of tweet text versus emojis.
• To evaluate different text sentiment analysis tools (LIWC, VADER, Hu Liu).

Main Methods:

• Analysis of over 7 million self-referencing and 18 million other tweets.
• Comparative sentiment analysis using Linguistic Inquiry and Word Count (LIWC), Valence Aware Dictionary and sEntiment Reasoner (VADER), and Hu Liu lexicon.
• Examination of temporal trajectories (circadian and circaseptan) of emotions in both text and emojis.

Main Results:

• Temporal emotion patterns show similarities and differences between self-referencing and other tweets.
• All sentiment analysis tools detected significant circadian and circaseptan patterns, with often corresponding shapes.
• Emoji-based emotion patterns occasionally differed from text-based patterns, suggesting complementary information.

Conclusions:

• Temporal emotion dynamics in tweets are influenced by content type (self-referential vs. other).
• Multiple sentiment analysis tools reveal consistent daily and weekly emotional rhythms.
• Emojis provide distinct, complementary insights into emotional expression, augmenting textual analysis for a richer understanding of well-being.