DeepADEMiner

DeepADEMiner is a system for extracting Adverse Drug Effect (ADE) (previously often described as Adverse Drug Reaction (ADR)) spans from user generated texts like tweets from Twitter or posts in Reddit. From a large collection of texts that contain drug/medication mentions, it extracts the spans where the ADEs are mentioned and normalizes them to their respective MedDRA preferred term identifiers. This work uses off-the-shelf named entity recognition (NER) and classification tools based on classical machine learning and newer deep learning frameworks to accomplish the above tasks.

Components of DeepADEMiner and illustration of tasks performed at each step

DeepADEMiner is available online as a demo and as a REST API.

Demo website: https://hlp.ibi.upenn.edu/deepademiner/
API documentation: https://bitbucket.org/pennhlp/deepademiner/src/docs/api.md

Code on Bitbucket: https://bitbucket.org/pennhlp/deepademiner/src
Data: https://bitbucket.org/pennhlp/deepademiner/src/data
[The source code and data for DeepADEMiner will be made available publicly on the date of peer-review article being published.]