Non-availability of well annotated and balanced datasets is considered as one of the major hurdles in analysing and extracting meaningful information from health-related tweets. Herein, we present transformer based deep learning binary classifiers for distinguishing the health related tweets for the three shared tasks 1a, 4 and 8 of the 6 th edition of SMM4H Workshop. We evaluate the different transformer based models viz. RoBERTa (for Task 1a & 4) and BioBERT (for Task 8), along with various dataset balancing techniques. We implement augmentation and sampling techniques so as to improve performance on the imbalanced datasets.

No abstract