United States

This paper explores the performance of multilingual models in the general domain on the clinical Question Answering (QA) task to observe their potential medical support for languages that do not benefit from the existence of clinically trained models. In order to improve the model&#39;s performance, we exploit multilingual data augmentation by translating an English clinical QA dataset into six other languages. We propose a translation pipeline including projection of the evidences (answers) into the target languages and thoroughly evaluate several multilingual models fine-tuned on the augmented data, both in mono- and multilingual settings. We find that the translation itself and the subsequent QA experiments present a differently challenging problem for each of the languages. Finally, we compare the performance of multilingual models with pretrained medical domain-specific English models on the original clinical English test set. Contrary to expectations, we find that monolingual domain-specific pretraining is not always superior to general-domain multilingual pretraining. The source code is available at https://github.com/lanzv/Multilingual-emrQA

NAACL 2025

When Multilingual Models Compete with Monolingual Domain-Specific Models in Clinical Question Answering

machine translation-based augmentation

clinical question-answering

multilingual models

EMNLP 2021 is planned to be a hybrid event in Punta Cana, Dominican Republic, with both on-site and fully virtual participation possible. The experience for on-site participants would closely approximate a normal pre-COVID *ACL conference, with 5-6 thematically organized parallel sessions and live Q/A and interactive discussion immediately after the talks. Presentations by virtual participants will be equitably interleaved with those of on-site participants, projected on the auditorium screens as if on-site, and also followed immediately by live Q/A and interactive discussion at a time during reasonable waking hours for the virtual presenter. For all participants, on-site and virtual, who are unable to attend a session due to either time-zone issues or because they are participating in another session live, talk recordings and slides will be available online at a minimum after the live presentation (and in many cases before as well), and questions may be submitted in advance on session-specific discussion boards and answered live in session with the usual visual aids if desired.

<iframe style="width:700px;height:400px" src="https://online.fliphtml5.com/ebtyf/ceby/" seamless="seamless" scrolling="no" frameborder="0" allowtransparency="true" allowfullscreen="true" ></iframe>

Please Note: The EMNLP registration system is not currently connected to the underline site as we are still in the process of building out EMNLP 2021. You will receive access instructions from underline the week of November 1st. 

Access is given only to EMNLP upon registration, if you have not registered please do so [here](https://2021.emnlp.org/registration).

Registered attendees will receive access the week of November 1st.

EMNLP 2021

EMNLP 2021 is planned to be a hybrid event in Punta Cana, Dominican Republic, with both on-site and fully virtual participation possible.

This paper explores the performance of multilingual models in the general domain on the clinical Question Answering (QA) task to observe their potential medical support for languages that do not benefit from the existence of clinically trained models. In order to improve the model's performance, we exploit multilingual data augmentation by translating an English clinical QA dataset into six other languages. We propose a translation pipeline including projection of the evidences (answers) into the target languages and thoroughly evaluate several multilingual models fine-tuned on the augmented data, both in mono- and multilingual settings. We find that the translation itself and the subsequent QA experiments present a differently challenging problem for each of the languages. Finally, we compare the performance of multilingual models with pretrained medical domain-specific English models on the original clinical English test set. Contrary to expectations, we find that monolingual domain-specific pretraining is not always superior to general-domain multilingual pretraining. The source code is available at https://github.com/lanzv/Multilingual-emrQA

workshop paper

### Welcome to 2025 Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics

Welcome to the 2025 meeting of the Nations of the Americas Chapter of the Association for Computational Linguistics! I am proud to help organize the first NAACL conference to carry the new name of our organization, one that emphasizes inclusion for all of the Americas. I am also pleased to welcome you to Albuquerque, New Mexico, a state whose unique blend of cultural influences will make for an excellent backdrop for NAACL 2025, especially with this year’s special theme on NLP in a Multicultural World. 
**[Continue reading...](https://drive.google.com/file/d/1jX-qGhqVSZZCIrAnJaz798pflu5Irrdn/view?usp=sharing)**

*- Colin Cherry, Google, NAACL 2025 General Chair* 

[![](https://assets.underline.io/markdown_image/1/image/b087f8a4dc5816d6e1a6514e59c59ac3.png)](https://drive.google.com/file/d/1T96GzPqObXrMln2BMByCSXTSizjTg69P/view?usp=sharing)

You need to log in with the email address you registered with. Access credentials have been sent to your email. 

Please be sure to check your spam and other email folders if you do not see an email confirmation right away.

Please log in to explore this event.

To access NAACL 2025 event page you are required to register. Please follow [**this link**](https://2025.naacl.org/registration/registration/) to register. Access will depend on your registration type.

Please register!

Welcome to 2025 Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics

In this paper, we address the challenge of patient-note identification, which involves accurately matching an anonymized clinical note to its corresponding patient, represented by a set of related notes. This task has broad applications, including duplicate records detection and patient similarity analysis, which require robust patient-level representations. We explore various embedding methods, including Hierarchical Attention Networks (HAN), three-level Hierarchical Transformer Networks (HTN), LongFormer, and advanced BERT-based models, focusing on their ability to process medium-to-long clinical texts effectively. Additionally, we evaluate different pooling strategies (mean, max, and mean_max) for aggregating word-level embeddings into patient-level representations and we examine the impact of sliding windows on model performance. Our results indicate that BERT-based embeddings outperform traditional and hierarchical models, particularly in processing lengthy clinical notes and capturing nuanced patient representations. Among the pooling strategies, mean_max pooling consistently yields the best results, highlighting its ability to capture critical features from clinical notes. Furthermore, the reproduction of our results on both MIMIC dataset and Necker hospital data warehouse illustrates the generalizability of these approaches to real-world applications, emphasizing the importance of both embedding methods and aggregation strategies in optimizing patient-note identification and enhancing patient-level modeling.

Comparing representations of long clinical texts for the task of patient-note identification

The massive size, continual growth, and technical jargon in biomedical publications make it difficult for laypeople to stay informed about the latest scientific advances, motivating research on lay summarization of biomedical literature. Large language models (LLMs) are increasingly used for this task. Unlike typical automatic summarization, lay summarization requires incorporating background knowledge not found in a paper and explanations of technical jargon. This study explores the use of MeSH terms (Medical Subject Headings), which represent an article's main topics, to enhance background information generation in biomedical lay summarization. Furthermore, we introduced a multi-turn dialogue approach that more effectively leverages MeSH terms in the instruction-tuning of LLMs to enhance the quality of lay summaries. The best model improved the state-of-the-art on the eLife test set in terms of the ROUGE-1 score by nearly 2%, with competitive scores in other metrics. These results indicate that MeSH terms can guide LLMs to generate more relevant background information for laypeople. Additionally, evaluation on a held-out dataset, one that was not used during model pre-training, shows that this capability generalizes well to unseen data, further demonstrating the effectiveness of our approach.

Towards Knowledge-Guided Biomedical Lay Summarization using Large Language Models

As digital health becomes more ubiquitous, people from different geographic regions are connected and there is thus a need for accurate language translation services. South Africa presents opportunity and need for digital health innovation, but implementing indigenous translation systems for digital health is difficult due to a lack of language resources. Understanding the accuracy of current models for use in medical translation of indigenous languages is crucial for designers looking to build quality digital health solutions. This paper presents a new dataset with audio and text of primary health consultations for automatic speech recognition and machine translation in South African English and the indigenous South African language of isiXhosa. We then evaluate the performance of well-established pretrained models on this dataset. We found that isiXhosa had limited support in speech recognition models and showed high, variable character error rates for transcription (26-70%). For translation tasks, Google Cloud Translate and ChatGPT outperformed the other evaluated models, indicating large language models can have similar performance to dedicated machine translation models for low-resource language translation.

Benchmarking IsiXhosa Automatic Speech Recognition and Machine Translation for Digital Health Provision

In this paper, we investigate using large language models to generate accessible lay summaries of medical abstracts, targeting non-expert audiences. We assess the ability of models like GPT-4 and LLaMA 3-8B-Instruct to simplify complex medical information, focusing on layness, comprehensiveness, and factual accuracy. Utilizing both automated and human evaluations, we discover that automatic metrics do not always align with human judgments. Our analysis highlights the potential benefits of developing clear guidelines for consistent evaluations conducted by non-expert reviewers. It also points to areas for improvement in the evaluation process and the creation of lay summaries for future research.

Mining Social Media for Barriers to Opioid Recovery with LLMs

Towards Understanding LLM-Generated Biomedical Lay Summaries

Large Language Models have made impressive progress in the medical field. In medical dialogue scenarios, unlike traditional single-turn question-answering tasks, multi-turn doctor-patient dialogue tasks require AI doctors to interact with patients in multiple rounds, where the quality of each response impacts the overall model performance. In this paper, we propose PERT to re-explore values of multi-turn dialogue training data after the supervised fine-tuning phase by integrating a prefix learning strategy, further enhancing the response quality. Our preliminary results show that PERT achieves notable improvements on gynecological data, with an increase of up to 0.22 on a 5-point rating scale.

Prefix-Enhanced Large Language Models with Reused Training Data in Multi-Turn Medical Dialogue

While increasing patients' access to medical documents improves medical care, this benefit is limited by varying health literacy levels and complex medical terminology. Large language models (LLMs) offer solutions by simplifying medical information. However, evaluating LLMs for safe and patient-friendly text generation is difficult due to the lack of standardized evaluation resources. To fill this gap, we developed MeDiSumQA. MeDiSumQA is a dataset created from MIMIC-IV discharge summaries through an automated pipeline combining LLM-based question-answer generation with manual quality checks. We use this dataset to evaluate various LLMs on patient-oriented question-answering. Our findings reveal that general-purpose LLMs frequently surpass biomedical-adapted models, while automated metrics correlate with human judgment. By releasing MeDiSumQA on PhysioNet, we aim to advance the development of LLMs to enhance patient understanding and ultimately improve care outcomes.

MeDiSumQA: Patient-Oriented Question-Answer Generation from Discharge Letters

Medical notes contain a wealth of information related to diagnosis, prognosis, and overall patient care that can be used to help physicians make informed decisions. However, like any other data sets consisting of data from diverse demographics, they may be biased toward certain subgroups or subpopulations. Consequently, any bias in the data will be reflected in the output of the machine learning models trained on them. In this paper, we investigate the existence of such biases in Danish medical notes related to three types of blood cancer, with the goal of classifying whether the medical notes indicate severe infection. By employing a hierarchical architecture that combines a sequence model (Transformer and LSTM) with a BERT model to classify long notes, we uncover biases related to demographics and cancer types. Furthermore, we observe performance differences between hospitals. These findings underscore the importance of investigating bias in critical settings such as healthcare and the urgency of monitoring and mitigating it when developing AI-based systems.

Bias in Danish Medical Notes: Infection Classification of Long Texts Using Transformer and LSTM Architectures Coupled with BERT

Health literacy enables individuals to navigate healthcare systems and make informed decisions. Plain language summaries (PLS) can bridge comprehension gaps by simplifying complex biomedical texts, yet their manual creation is both time-consuming and challenging. This study advances the field by (1) constructing a novel corpus of paired technical and plain language texts from medical trial libraries, (2) developing machine learning classifiers to rapidly identify plain language features, and (3) establishing a multi-dimensional evaluation framework that integrates computational metrics with human expertise. We iteratively optimized prompts for diverse large language models (LLMs)---including GPT models, Gemini 1.5, DeepSeek-R1, and Llama-3.2---to generate PLS variants aligned with domain-specific guidelines. Our classifier achieved 97.5% accuracy in distinguishing plain from technical language, and the generated summaries demonstrated high semantic equivalence to expert-written versions.

Bridging the Gap in Health Literacy: Harnessing the Power of Large Language Models to Generate Plain Language Summaries from Biomedical Texts

Clinical documents are essential to patient care, but their complexity often makes them inaccessible to patients. Large Language Models (LLMs) are a promising solution to support the creation of lay translations of these documents, addressing the infeasibility of manually creating these translations in busy clinical settings. However, the integration of LLMs into medical practice in Germany is challenging due to data scarcity and privacy regulations. This work evaluates an open-source LLM for lay translation in this data-scarce environment using datasets of German synthetic clinical documents and real tumor board protocols. The evaluation framework used combines readability, semantic, and lexical measures with the G-Eval framework. Preliminary results show that zero-shot prompts significantly improve readability (e.g., FREde: 21.4 → 39.3) and few-shot prompts improve semantic and lexical fidelity. However, the results also reveal G-Eval's limitations in distinguishing between intentional omissions and factual inaccuracies. These findings underscore the need for manual review in clinical applications to ensure both accessibility and accuracy in lay translations. Furthermore, the effectiveness of prompting highlights the need for future work to develop applications that use predefined prompts in the background to reduce clinician workload.

Downloads

Next from NAACL 2025

Comparing representations of long clinical texts for the task of patient-note identification

Similar lecture

Clustering Monolingual Vocabularies to Improve Cross-Lingual Generalization

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from NAACL 2025

Comparing representations of long clinical texts for the task of patient-note identification

Similar lecture

Clustering Monolingual Vocabularies to Improve Cross-Lingual Generalization

Downloads