Brundlefly at SemEval-2016 Task 12: Recurrent Neural Networks vs. Joint Inference for Clinical Temporal Information Extraction

TL;DR

This paper compares recurrent neural networks and joint inference methods for extracting and relating clinical temporal information in medical notes, showing joint inference with structured prediction outperforms RNNs for entity extraction.

Contribution

It introduces a joint inference-based structured prediction approach for clinical temporal entity extraction and relation classification, demonstrating its superiority over vanilla RNNs in this domain.

Findings

Structured prediction outperforms RNNs for entity extraction.

Date canonicalization and distant supervision improve relation classification.

Limited gains due to small training dataset.

Abstract

We submitted two systems to the SemEval-2016 Task 12: Clinical TempEval challenge, participating in Phase 1, where we identified text spans of time and event expressions in clinical notes and Phase 2, where we predicted a relation between an event and its parent document creation time. For temporal entity extraction, we find that a joint inference-based approach using structured prediction outperforms a vanilla recurrent neural network that incorporates word embeddings trained on a variety of large clinical document sets. For document creation time relations, we find that a combination of date canonicalization and distant supervision rules for predicting relations on both events and time expressions improves classification, though gains are limited, likely due to the small scale of training data.