How to handle noisy sentiment data in micro climate studies for data analysis?

[Ella.W]

I'm working on a personal research project analyzing the correlation between local weather patterns and the tone of community social media posts over a year. The data analysis is becoming overwhelming because the sentiment data is so noisy. I'm curious if anyone has tried similar micro-scale socio-climatic studies and what methods you found most effective for cleaning and interpreting such messy, real-world datasets.