Sentiment analysis (supervised and unsupervised polarity classification) of Twitter data about the Albania 2019 earthquake

This database contains the comparison of results between the supervised and unsupervised polarity classification of the tweets related to the 2019 Albania earthquake. The database was constructed with the aim to test the accuracy of the algorithm developed by MonkeyLearn for polarity classification.

675 Tweets with the hashtags: #Albania #AlbanianEarthquake #albanianearthquake from the 26th November 2019 to the 3rd February 2020 were collected by the third-party vendor: TweetBinder. The social media department of Newcastle University provided us with 1001 tweets with the hashtags: #Albania collected from the 31st January to the 2nd February 2020. After removing repeated tweets from the database, we obtained a dataset made up of 255 original tweets. This database only contains text data of original tweets (no retweets). Attributes and data contained:

-N: Number of the tweet

-Tweet: Text data

-ML-Classification: Unsupervised polarity classification performed by the algorithm developed by MonkeyLearn.

-Confidence: Percentage of trust that the predicted polarity is right

-RB-Classification: Supervised polarity classification performed by experts.

-Accuracy: Coincidence indicated by 1 in polarity predicted by the unsupervised classification and the supervised classification. No coincidence is indicated by 0.

-TPt: true positive

-FPtNg: misclassified as Positive then False positive, when it is negative

-FPtNt: misclassified as Positive then False positive , when it is Neutral

-TNg: true negative

-FNgPt: missclasified as Negative then False negative when it is Postive.

-FNgNt: missclasified as Negative then False negative when it is Neutral.

-TNt: true neutral

-FNt Pt: missclasified as Neutral, then False-Neutral when it is Positive

-FNtNg: and missclasified as Neutral, then False-Neutral, when it is negative.