Yep, that's what i did. Every 15 min I would measure the sentiment of given topics, normalize the messy data and plot them. you can overlay this against a time

Yep, that's what i did. Every 15 min I would measure the sentiment of given topics, normalize the messy data and plot them. you can overlay this against a time series of price data. Twitter also has a historic tweets api. It's more work, but you can backtest as well.