gsuriano / twittersentimentanalyses

Sentiment analyses task using traditional ML techniques

Twitter Sentiment Analysis

Repository for the Data Science Lab course final project.

As final project for the Data Science Lab course at Politecnico di Torino, me and my teammate for this project, Pietro Montresori, have developed an algorithm which performs Sentiment Analysis.

The used data are in the publicly available sentiment140 dataset, and as in most Data Science competitions were splitted in a public part, upon which training the model, and a private part, to do predictions to finally be evaluated. Up to 4 points over 31 were assigned with respect to a leaderboard defined in the proprietary Politecnico platform, DSLE.

The approach I and my teammate have proposed is basically rooted on text: even though informations about the user logically can help the model predicting the sentiment, this information were not part of the dataset and, for the sake of generality (and re-usability) we then preferred to direct our efforts into building a model who takes as input for the sentiment prediction text only.

More details related to our approach are presented in the paper associated with this project.