The increasing availability of multimodal data from electronic health records (EHR) has paved the way for deep learning methods to improve diagnosis accuracy. However, deep learning models are data-driven, requiring large-scale datasets to achieve high generalizability. Inspired by how human experts leverage reasoning for medical diagnosis, we propose CARER, a novel health risk prediction framework, that enhances deep learning models with clinical rationales derived from medically proficient Large Language Models (LLMs). In addition, we provide a cross-view alignment loss which aligns the “local” view from the patient’s health status with the “global” view from the external LLM’s clinical reasoning to boost the mutual feature learning. Through extensive experiments on two predictive tasks using two popular EHR datasets, our CARER’s significantly exceeds the performance of state-of-the-art models by up to 11.2%, especially in improving data efficiency and generalizability. Our code is available at https://github.com/tuandung2812/CARER-EMNLP-2024

The aspect-based sentiment analysis (ABSA) consists of two conceptual tasks, namely an aspect extraction and an aspect sentiment classification. Rather than considering the tasks separately, we build an end-to-end ABSA solution. Previous works in ABSA tasks did not fully leverage the importance of syntactical information. Hence, the aspect extraction model often failed to detect the boundaries of multi-word aspect terms. On the other hand, the aspect sentiment classifier was unable to account for the syntactical correlation between aspect terms and the context words. This paper explores the grammatical aspect of the sentence and employs the self-attention mechanism for syntactical learning. We combine part-of-speech embeddings, dependency-based embeddings and contextualized embeddings (e.g. BERT, RoBERTa) to enhance the performance of the aspect extractor. We also propose the syntactic relative distance to de-emphasize the adverse effects of unrelated words, having weak syntactic connection with the aspect terms. This increases the accuracy of the aspect sentiment classifier. Our solutions outperform the state-of-the-art models on SemEval-2014 dataset in both two subtasks.