This is a modified version of getting past tweets from Jefferson Henrique (https://github.com/Jefferson-Henrique/GetOldTweets-python)
python3 GetOldTweets3.py
We need to remove mention, website, non-ASCII, keywords before analysis
python3 dataClean.py
TweetAnalysis.ipynb
Here I am using the nltk package to do the sentiment analysis
I am using the 200,000 for training the LDA model and analysis the whole dataset. The dataset can be clustered into 27 topics. Topics are shown below:
Topic | |
---|---|
0 | hand, wash, tw, prisoner, clinical_trial, officer, writes, water, changing, soap |
1 | india, warned, trump, hart_island, oil, loved_one, latino, nyt, forever, changed |
2 | resource, join, pandemic, crisis, student, help, free, curve, webinar, health |
3 | like, people, know, shit, got, thing, going, gonna, day, think |
4 | dr_fauci, fauci, florida, texas, plasma, poverty, trump, saved, pollution, laid |
5 | stay, stayhome, home, nyc, safe, igshid, quarantine, time, socialdistancing, new |
6 | relief, worker, fund, help, pandemic, support, health, healthcare, donate, need |
7 | tiger, homeless, worker, essential, bronx_zoo, positive, test, mta, shelter, new |
8 | spain, grocery, initiative, threatens, hardest_hit, brief, pause, singapore, weekly, pledge |
9 | sign, petition, pandemic, trump, intelligence, month, spread, crisis, american_moneyforthepeople, projection |
10 | contact_tracing, mom, reopen, survivor, utm_source, minority, campaign, lost, project, shutdown |
11 | mask, face, wearing, wear, people, igshid, glove, like, articleshare, html_referringsource |
12 | testing, vaccine, trial, voting, site, launch, mail, vote, record, south_korea |
13 | black, people, rate, data, death, american, dying, higher, african, killing |
14 | case, death, new, york, death_toll, number, state, confirmed, city, total |
15 | update, live, briefing, mayor, blasio, latest, watch, news, new, task_force |
16 | prison, jail, inmate, passed_away, pandemic, release, charity, housing, rent, crisis |
17 | patient, drug, hydroxychloroquine, treatment, hospital, treat, doctor, supply, survey, study |
18 | tracking, privacy, apple_google, equipment, scam, google, technology, track, misinformation, news |
19 | cat, wisconsin, pet, dog, russia, easter_sunday, election, real_estate, animal, warns |
20 | trump, china, president, american, hoax, response, people, donald, biden, democrat |
21 | map, icu, amazon, feature, defeat, ea, challenging, bailout, democracy, cc |
22 | ventilator, frontline, fuck, sport, hero, peak, season, game, celebrity, fan |
23 | gate, inequality, nursing_home, pandemic, reveals, recovering, startup, crisis, utm_medium, success |
24 | passover, flight, airline, refund, tracker, cousin, payment, cancel, ease, pastor |
25 | test, recovered, flu, tested, antibody, people, symptom, immigrant, testing, know |
26 | easter, boris_johnson, honor, celebrate, navigate, veteran, church, chloroquine, tribute, magazine |
Combine the topics and sentiment results.
The hot topics in tweet are topics 2, 3, 5, 6,14, 20.