Tweet sentiment Classification

Here, I explore some machine learning techniques for handling text data.

Process

1. Literature

My research work at this point is in graph-based networks for fact verification rather than sentiment analysis, so I decided to look through the relevant literature and familiarize myself.

The papers are documented in "literature".

2. Exploration and development of models

Using a Jupyter Notebook to document my findings and progress, I clean and explore the dataset and implement a simple logistic regresison model using TF-IDF to encode the text, and a simple pre-trained BERT model for comparison.

Both models have decent performance, with classification accuracy, precision, recall and F1-scores of about 90%. These metrics have been chosen to give a holistic picture of classification performance.

How to

To get this up and running easily, I recommend installing Anaconda.