This repository attempts to apply sentiment analysis on football commentary data that is downloaded from Kaggle here.
The main objective is to:
Classify which teams have the most positive and negative sentiment as revealed by the football commentary.
- This is treated as a proxy for how 'good' a team is because intuitively speaking, the more posiive sentiment commentators express regarding a team in a game, the more likely that team is performing well and henceforth, over a number of games played that's captured in the dataset, this provides an indicator of how good a team is.
In this way, we can compare the outputs with our own common knowledge of which teams are good (indeed there is scope to map this against data on league standings to provide an objective way to add labels and turn this unsupervised problem into a supervised one.)