Clinical Text Classification with Rule-based Features and Knowledge-guided Convolutional Neural Networks

TL;DR

This paper introduces a novel method combining rule-based features and knowledge-guided deep learning, specifically CNNs with UMLS embeddings, to improve clinical text classification, outperforming existing approaches.

Contribution

It presents a new hybrid approach integrating rule-based features with CNNs and UMLS embeddings for disease classification in clinical texts.

Findings

Outperforms state-of-the-art methods on i2b2 obesity challenge

Effective use of trigger phrases for few-shot class prediction

Combines rule-based and deep learning techniques successfully

Abstract

Clinical text classification is an important problem in medical natural language processing. Existing studies have conventionally focused on rules or knowledge sources-based feature engineering, but only a few have exploited effective feature learning capability of deep learning methods. In this study, we propose a novel approach which combines rule-based features and knowledge-guided deep learning techniques for effective disease classification. Critical Steps of our method include identifying trigger phrases, predicting classes with very few examples using trigger phrases and training a convolutional neural network with word embeddings and Unified Medical Language System (UMLS) entity embeddings. We evaluated our method on the 2008 Integrating Informatics with Biology and the Bedside (i2b2) obesity challenge. The results show that our method outperforms the state of the art methods.