Social network users provide a wealth of hand written weather reports and observations. These reports often show that a human experiences weather in a way that is not conveyed well through automated weather observations. Automatically classifying these hand written weather reports can be valuable for use in automated weather analysis tools. However classifying hand written weather reports presents many unique challenges in the initial classification as well as the verification of resulting classified reports.

This presentation will cover using Natural Language Processing and Sentiment Analysis techniques to classify weather reports and present an automated system for classifying reports in real-time. This presentation will also cover verification of the classified reports and discuss issues relating to poor quality reports and irrelevant reports.