Closed marcdotson closed 5 years ago
Note we might need to include a separate column as "maybe" for tweets that aren't obviously complaints.
Here is a link to all of our labeled tweets. 4960 labeled tweets in total.
Let's put this data on the repo.
Do we need to build a better test data set? Let's get it in the Data directory. @AdrielC @dallin-cardon
We need to create a test dataset by labeling 5000 tweets as complaints. We'll use this to validate and compare different classification techniques.
Adriel has the sheet -- just post a link here for the shared document.