Drug-drug interaction (DDI) has the potential to cause significant harm to the patient. In this poster, we propose a transfer learning framework with pre-trained BERT model for extracting DDI information from Structured Product Labeling (SPL) documents. The results of the evaluation on the Text Analysis Conference (TAC) 2018 DDI track datasets show our model outperforms one of the state-of-the-art models, and the integration of the pre-trained model can improve the performance of DDI extraction tasks.

Learning Objective: After participating in this session, the learner should be better able to:
1. Understand the current issues in automate DDI extraction, and be able to discuss and recommend several natural language processing (NLP) approaches for extracting DDI information from SPL documents.
2. Learn the process of transfer learning including pre-training a model on large amounts of unlabeled texts and transferring the learned general language knowledge to a specific NLP task through fine-tuning.
3. Understand the pre-trained model can improve performances on DDI extraction tasks because the context can be encoded in its textural representations.


Yuqing Mao (Presenter)

Kin Wah Fung, NLM
Dina Demner-Fushman, NLM

Presentation Materials: