![]() |
Language Technology Programming Competition 2025 | ||||||||||||||||||||
|
2025 Shared Task DescriptionBasic Task DescriptionBackgroundActive adverse event surveillance monitors Adverse Drug Events (ADEs) across various data sources, such as electronic health records, medical literature, social media, and search engine logs. Three key NLP tasks—forming a typical monitoring pipeline—are:
GoalThe goal of this shared task is to focus on the third task described above—--normalizing ADE mentions to MedDRA. That is, participants are expected to develop a system that takes as input a user-written post with identified adverse drug event descriptions and outputs a ranked list of the most relevant MedDRA terms for each adverse drug event. The performance of the systems will be evaluated using the Accuracy@n score. Specifically, for each test instance (i.e., one adverse drug event), we check whether the ground truth label appears among the top n results in the system's predictions. We will use Accuracy@1 as the primary evaluation metric to rank participant systems. Additionally, Accuracy@5 and Accuracy@10 will be calculated as reference metrics. Data Files and SubmissionWe will use CodaBench for this year's competition (ALTA Shared Task 2025). The details about data formats and the submission will be provided in the competition website. Important Dates
|
||||||||||||||||||||