vcuspinera / Canada_response_covid

Sentiment analysis of tweets after Government announcement of policies as response of Covid-19
MIT License
0 stars 0 forks source link

TODO list with pending tasks #3

Closed vcuspinera closed 3 years ago

vcuspinera commented 3 years ago

TODO

cuspime commented 3 years ago
cuspime commented 3 years ago

I think this issues way of doing things is lovely btw. Much better than what I did. Lol, soz

cuspime commented 3 years ago

as I said, Checkout the end of eda-Copy1, I did some analysis.

vcuspinera commented 3 years ago

@cuspime I will look into the eda-Copy1; you can also update the original eda file with the time series analysis. I just finished the changes on the Heatmap following your recommendation of Normalizing the Moving Average of tweets using min-max values of each twitter account, as mentioned in the issue #4

vcuspinera commented 3 years ago

Just add to the Sentiment Analysis Jupyter Notebook the polarity and subjectivity plots by Canadian Government Twitter account, as mentioned in issue #6.

vcuspinera commented 3 years ago

I looked at the eda-Copy1 and I like a lot the Time Series analysis, it really add up to the project Leo @cuspime.

I realize that you add this work to the Exploratory Data Analysis (EDA), but as you are using the polarity and the final dataset tweets_db_sentiment.json, I think this work should be added in the main Sentiment Analysis of the Repo.

So, I will add the analysis to the sentiment_analysis.ipynb file. Also we should delete the eda-Copy1.

cuspime commented 3 years ago

Lovely! I'm happy you liked it. Maybe we can also do something similar with the amount of tweets? I mean I don't think it would be as central as an analysis but it might...

cuspime commented 3 years ago

Cleaning ads: We can spot some ads by running

pd.set_option('display.max_colwidth', None)
df_tot[df_tot[['account', 'lang', 'sourceLabel','replyCount', 'retweetCount', 'likeCount', 
               'username', 'tweet', 'polarity',
       'subjectivity']].duplicated()][['account', 'lang', 'replyCount', 'retweetCount', 'likeCount','username', 'tweet', 'polarity',
       'subjectivity']].sort_values(by=[ 'retweetCount', 'likeCount','replyCount'], ascending=False)
cuspime commented 3 years ago

Let's check the hashtags on the 10th and on the 25th

vcuspinera commented 3 years ago

In the next table I am sharing the ranking of Twitter Trends in Canada, on March 10th and 25th of 2020, at 12:00 and 18:00 hr.

This information was retrieved from the getdaytrends webpage.

Twitter Trends in Canada:

Ranking Mar-10, 12:00 Mar-10, 18:00 Mar-25, 12:00 Mar-25, 18:00
1 #ThankYouNamjoon #MAR10Day Prince Charles CERB
2 #TuesdayMotivation #BernieSurge #Mixtape_OnTrack Quarantine Act
3 #MAR10Day #TuesdayThoughts #Mixtape_바보라도알아 Prince Charles
4 Chuck Norris Warzone #WednesdayMotivation #IKnewIHadCabinFeverWhen
5 #oriandthewillofthewisps Ollie #FreeJamesWoods #WednesdayMotivation
6 #CoreyFeldman anne marie #PrayTogether kamal
7 italy AR-14 easter Anime North
8 Hannah Ann #newcitiesglobalgoals Liberals #startupchats
9 Gobert #MattGaetzIsAQuarantinedTool parliament #boycottTimHortons
10 Sydow Chuck Norris Scott Reid Bill C-13
11 Happy Holi Ricky 1 in 5 Canadians #ButterflyXIUDay
12 CPAC Muskrat Falls Conservatives Royal Assent
13 Norm Morgan Rielly LCBO Connor Bedard
14 Peter Horizon Zero Dawn Swizz Stuart Gordon
15 Pearl Jam Bob Rae Tory Lanez Canada Goose
16 mccaw Ivy League Dave Stieb Our Father
17 Strange Brew NFL 2K Pierre Bob and Doug McKenzie
18 Utah Marleau Ann Coulter Alex Ottley
19 Gamestop Charlie Sheen my 2,400 senat
20 Ibaka 2nd Amendment Dick Pound Lesley
21 #HowNotToGetArrested Éric Salvail Olympics Sikhs
22 #PearlJam hedman Bloc Communes
23 #TheBachelor Schlatter Albert Uderzo Romanov
24 #COVID2019 wheel of fortune timbo Our Government
25 #InternationalWomensDay Madden #CoronavirusLockdown Princess Diana
vcuspinera commented 3 years ago

⚠️ After some discussions we realize that the first announcement made by the Prime Minister of Canada was done on March 11th, which make sense. From this point, we download tweets from February 1st until April 30th of 2020, change the announcement date to March 11th, and re-run every analysis we made before

vcuspinera commented 3 years ago

I assume that Polarity and Subjectivity equals to zero is related with tweets that use words that probably are not in the TextBlob dictionary, or have words that are not subjective nor positive or negative polarity, so it assigns values as zero. Some of these tweets as examples:

vcuspinera commented 3 years ago

We finished the pending tasks and will versioning the project as closing.