Skip to content
Social Sense

Twitter · Victoria · Feb-Jul 2022

How each suburb sounded on Twitter

The team's MPI processors read the University's Twitter corpus line by line, scored every tweet from 1 (extremely negative) to 9 (extremely positive) and matched its place name to a 2021 ABS suburb. CouchDB MapReduce views then summed the scores per suburb. This is that map, rebuilt from the saved view results.

Geotagged tweets in Victoria
719,336
of 2,418,617 Australia-wide
Victorian suburbs with tweets
1,085
ABS SAL 2021
Scored neutral (5)
48%
the single most common score
Tweets processed in 2023
37,823,414
15.9 GB of the 57 GB corpus

Explore

The suburb map

Tweets
Colour by
Minimum tweets per suburb≥ 1
Zoom to
Find

Loading map…
Average sentiment score (1 = very negative, 9 = very positive)
≤ 3.5 more negative5.5more positive ≥ 7.5
No tweets at this threshold

Overall

Mostly neutral, leaning positive

Across the whole corpus the scores pile up at 5 and the positive side is heavier than the negative, the same "slightly right-skewed" shape the team described. Scores come from NLTK's VADER compound score in fixed 0.2-wide bands; you can try the exact pipeline on the pipeline page.

All geotagged tweets

2,418,617 tweets with a matched suburb

1 · Extremely Negative: 39,726 (1.6%)1.6%12 · Very Strongly Negative: 80,238 (3.3%)3.3%23 · Strongly Negative: 132,745 (5.5%)5.5%34 · Negative: 120,487 (5.0%)5.0%45 · Neutral: 1,158,765 (47.9%)48%56 · Positive: 187,854 (7.8%)7.8%67 · Strongly Positive: 321,908 (13.3%)13%78 · Very Strongly Positive: 241,710 (10.0%)10.0%89 · Extremely Positive: 135,184 (5.6%)5.6%9← NegativeNeutralPositive →
Sentiment score distribution of all geotagged tweets
ScoreCountShare
1 (Extremely Negative)397261.6%
2 (Very Strongly Negative)802383.3%
3 (Strongly Negative)1327455.5%
4 (Negative)1204875.0%
5 (Neutral)115876547.9%
6 (Positive)1878547.8%
7 (Strongly Positive)32190813.3%
8 (Very Strongly Positive)24171010.0%
9 (Extremely Positive)1351845.6%

Rankings

Busiest, happiest and gloomiest suburbs

Most tweets

  1. 1Melbourne587,004 · avg 5.54
  2. 2Geelong21,729 · avg 5.49
  3. 3Ballarat Central18,929 · avg 5.19
  4. 4Castlemaine8,305 · avg 5.65
  5. 5Bendigo4,920 · avg 5.77
  6. 6Hillside (Melton - Vic.)3,404 · avg 5.72
  7. 7Lara3,236 · avg 5.66
  8. 8Traralgon3,108 · avg 5.47
  9. 9Koroit3,004 · avg 5.53
  10. 10Logan2,885 · avg 5.57

Most positive (≥ 100 tweets)

  • Flemington152 tweets · 6.5995% CI 6.32–6.87
  • Bacchus Marsh320 tweets · 6.5895% CI 6.38–6.78
  • Werribee185 tweets · 6.4895% CI 6.22–6.74
  • Leongatha358 tweets · 6.4695% CI 6.29–6.64
  • Halls Gap309 tweets · 6.4195% CI 6.22–6.59
  • Mount Lonarch341 tweets · 6.4095% CI 6.20–6.61

Most negative (≥ 100 tweets)

  • Flinders (Vic.)147 tweets · 4.7895% CI 4.45–5.11
  • Drouin369 tweets · 4.9395% CI 4.76–5.11
  • Metung146 tweets · 4.9595% CI 4.75–5.16
  • Creswick715 tweets · 5.0295% CI 4.87–5.16
  • Camperdown (Vic.)144 tweets · 5.0395% CI 4.71–5.36
  • Sea Lake368 tweets · 5.0395% CI 4.95–5.11

Intervals are t-based 95% intervals for each suburb's mean, from the count, sum and sum of squares the CouchDB views kept; tweets are treated as independent, so they are optimistic. Within each list every interval overlaps every other, so the order inside a list is not a finding. None of the 102 suburbs with at least 100 tweets has an interval wholly below the neutral 5. Place names come from Twitter's self-reported place field, matched to suburbs by the team's n-gram lookup; a tweet tagged "Melbourne, Victoria" lands in the Melbourne CBD suburb, which is why it dominates every ranking.