Essentially the user was dissatisfied with the results of the sentiment analysis and had to manually recode some of the responses and was curious if we could have a sort of “improvement feedback loop” baked into the tool.

Step 1: Do the sentiment analysis

Step 2: User goes into the sentiment analysis and identifies "wrong" calculations (i.e., "I never had any problems" should be coded as +1 instead of -1, for example)

Step 3: Sentiment analysis is re-run using the user input from Step 2 and gets a little more accurate. 

Step 4: Repeat as needed