Problem

The previous Xenophobia Meter Project's sentiment analysis tool uses a very naive, rule-based system of Xenophobic Sentiment Classification. This causes inaccuracies in their classification system. They would like to improve their classification abilities with a modern, NLP-based solution.

Solution

We obtained a model that averaged an accuracy of 74% with about ±5% across all accuracy metrics. This percentage is expected to increase with larger and more balanced datasets.