Table 2.
Twitter engagement of food security tweets with different highest-probability topics created through topic modeling and sentiment analysisa (superscripted letters denote statistical significance).
|
|
Engagement total, median (IQR) | Engagement in 2019 (n=10,562), median (IQR) | Engagement in 2020 (n=15,302), median (IQR) | Engagement in 2021 (n=12,206), median (IQR) | ||||||
| Engagement (sum of likes, replies, quotes, and retweets) | 11 (2-165) | 10 (2-141)b | 11 (2-123)b | 15 (3-270)c | ||||||
| Sentiment | ||||||||||
|
|
Very negative | 18 (2-205)d | 20.5 (4-560)b | 38 (4-126)b | 10 (1-209)b,e | |||||
|
|
Negative | 22 (3-238)d | 25 (4-244)b | 28 (4-381)b | 13.5 (2-128)b | |||||
|
|
Neutral | 11 (2-304)f | 7 (2-86)c | 9 (2-99)c | 34 (3-930)c | |||||
|
|
Positive | 9 (2-100)g | 7 (2-102)c | 7 (2-46)h | 13 (3-247)h | |||||
|
|
Very positive | 6 (2-22)i | 4 (2-9)h | 9 (3-94)c,h | 5 (2-14)e | |||||
| Topics and predominant sentiment | ||||||||||
|
|
Global production | |||||||||
|
|
|
Positive | 5 (2-14)d | 4 (1-10)b | 6 (2-18)b | 5 (2-15)b | ||||
|
|
Food insecurity and health | |||||||||
|
|
|
Negative | 7 (2-25)f | 7 (2-19)c | 8 (2-34)c | 6 (2-22)b | ||||
|
|
Use of food banks | |||||||||
|
|
|
Positive and neutral | 37 (3-501)g | 85 (5-1715)e,h | 25 (2-361)e,h | 48 (3-501)c | ||||
|
|
Giving to food banks | |||||||||
|
|
|
Positive | 8 (2-79)i | 8 (2-58)j,k | 8 (2-73)c | 7 (2-131)h | ||||
|
|
Family poverty | |||||||||
|
|
|
Negative | 111 (5-2167)l | 285 (11-4033)m | 125 (4-2214)j,n | 42 (3-1288)c | ||||
|
|
Food relief provision | |||||||||
|
|
|
Positive | 12 (3-171)o | 4 (1-10)b,c | 6 (2-18)b | 51 (5-930)c | ||||
|
|
Global food insecurity | |||||||||
|
|
|
Negative | 13 (2-129)i,o | 6 (1-21)b,c,j | 38 (4-238)h | 6 (1-86)b | ||||
|
|
Climate change | |||||||||
|
|
|
Positive | 142 (14-581)l | 78 (5-244)h,n | 25 (3-345)e,h,j,m | 200 (43-609)e | ||||
|
|
Australian food insecurity | |||||||||
|
|
|
Negative | 19 (3-141)o | 23 (2-141)k,n | 18 (3-121)e,m | 11 (3-247)h | ||||
|
|
Human rights | |||||||||
|
|
|
Positive | 126 (3-1053)g,l | 1053 (5-1053)e,m | 381 (3-8509)n | 4 (1-127)b,h | ||||
aPredominant sentiment refers to the sentiment with the highest proportion for each topic, as shown in Figure 6 and Table S4 in Multimedia Appendix 3. P<.001 Kruskal-Wallis test for differences between topic and year and differences between sentiment categories.
b-oValues within topic overall, sentiment overall, and topic and sentiment by each year with different superscript letters are significantly different from each other using the post hoc Dunn test and Bonferroni correction.