The designed binary classifier should detect tweets where Twitter users self-declare changing their medication treatments, regardless of being advised by a health care professional to do so. Such changes are, for example, not filling a prescription, stopping a treatment, changing a dosage, forgetting to take the drugs, etc. This task is the first step toward detecting patients non-adherent to their treatments and their reasons on Twitter. The data consists of two corpora: a set of tweets and a set of drug reviews from WebMD.com. Negative and positive reviews are naturally balanced whereas positive and negative tweets are naturally imbalanced. Each set is split into a training, a validation, and a test subset. The participants will be given the training and validation subsets for both corpora and evaluated on both test sets independently. Participants are expected to submit their predictions for both test sets. This year, we will add in the test sets additional reviews and tweets as decoys to avoid manual corrections of the predicted labels. Evaluation script, annotation guidelines, and baseline code will be provided to registered participants.
- Training data: 5,898 Tweets / 10,378 Reviews
- Validation data: 1,572 Tweets / 1,297 Reviews
- Test data: 2,360 Tweets / 1,297 Reviews
- Evaluation metric: F1-score for the change class
Register your team here : https://forms.gle/1qs3rdNLDxAph88n6
Link to Codalab : Available Feb 1 2021
Contact information: Davy Weissenbacher (email@example.com)