Progress in biomedical Named Entity Recognition (NER) and Entity Linking (EL) is currently hindered by a fragmented data landscape, a lack of resources for building explainable models, and the limitations of semantically-blind evaluation metrics. To address these challenges, we present MedPath, a large-scale and multi-domain biomedical EL dataset that builds upon nine existing expert-annotated EL datasets. In MedPath, all entities are 1) normalized using the latest version of the Unified Medical Language System (UMLS), 2) augmented with mappings to 62 other biomedical vocabularies and, crucially, 3) enriched with full ontological paths -- i.e., from general to specific -- in up to 11 biomedical vocabularies. MedPath directly enables new research frontiers in biomedical NLP, facilitating training and evaluation of semantic-rich and interpretable EL systems, and the development of the next generation of interoperable and explainable clinical NLP models.

Retrieval-augmented learning based on radiology reports has emerged as a promising direction to improve performance on long-tail medical imaging tasks, such as rare disease detection in chest X-rays. Most existing methods rely on comparing high-dimensional text embeddings from models like CLIP or CXR-BERT, which are often difficult to interpret, computationally expensive, and not well-aligned with the structured nature of medical knowledge. We propose a novel, ontology-driven alternative for comparing radiology report texts based on clinically grounded concepts from the Unified Medical Language System (UMLS). Our method extracts standardised medical entities from free-text reports using an enhanced pipeline built on RadGraph-XL and SapBERT. These entities are linked to UMLS concepts (CUIs), enabling a transparent, interpretable set-based representation of each report. We then define a task-adaptive similarity measure based on a modified and weighted version of the Tversky Index that accounts for synonymy, negation, and hierarchical relationships between medical entities. This allows efficient and semantically meaningful similarity comparisons between reports. We demonstrate that our approach outperforms state-of-the-art embedding-based retrieval methods in a radiograph classification task on MIMIC-CXR, particularly in long-tail settings. Additionally, we use our pipeline to generate ontology-backed disease labels for MIMIC-CXR, offering a valuable new resource for downstream learning tasks. Our work provides more explainable, reliable, and task-specific retrieval strategies in clinical AI systems, especially when interpretability and domain knowledge integration are essential. Our code is available at https://github.com/Felix-012/ontology-concept-distillation

Self-disclosure is important to help us feel better, yet is often difficult. This difficulty can arise from how we think people are going to react to our self-disclosure. In this workshop paper, we briefly discuss self-disclosure to conversational user interfaces (CUIs) in relation to various social cues. We then, discuss how expressions of uncertainty or representation of a CUI's reasoning could help encourage self-disclosure, by making a CUI's intended "theory of mind" more transparent to users.

Electronic health record (EHR) systems contain a wealth of multimodal clinical data including structured data like clinical codes and unstructured data such as clinical notes. However, many existing EHR-focused studies has traditionally either concentrated on an individual modality or merged different modalities in a rather rudimentary fashion. This approach often results in the perception of structured and unstructured data as separate entities, neglecting the inherent synergy between them. Specifically, the two important modalities contain clinically relevant, inextricably linked and complementary health information. A more complete picture of a patient's medical history is captured by the joint analysis of the two modalities of data. Despite the great success of multimodal contrastive learning on vision-language, its potential remains under-explored in the realm of multimodal EHR, particularly in terms of its theoretical understanding. To accommodate the statistical analysis of multimodal EHR data, in this paper, we propose a novel multimodal feature embedding generative model and design a multimodal contrastive loss to obtain the multimodal EHR feature representation. Our theoretical analysis demonstrates the effectiveness of multimodal learning compared to single-modality learning and connects the solution of the loss function to the singular value decomposition of a pointwise mutual information matrix. This connection paves the way for a privacy-preserving algorithm tailored for multimodal EHR feature representation learning. Simulation studies show that the proposed algorithm performs well under a variety of configurations. We further validate the clinical utility of the proposed algorithm in real-world EHR data.

Recently, I needed to book a lunch meeting. To help coordinate, I asked Amy to assist and cc'd her on the email. "Amy," I wrote, "please help us find a time to meet. Let's plan for sushi at Tokyo Express on Spear Street." Amy looked at my calendar, found an open time suitable for everyone invited, and booked the meeting.

Conversational user experiences, in the form of chatbots and voice interfaces, are overtaking many of the traditional ways in which we interact with machines. Since the rise of computers, human-machine interfaces typically had some form of Graphical User Interface (GUI) which enabled direct (if limited) interaction with devices and their programs, for instance via software installs, mobile apps, and web-based applications such as Software as a Service (SaaS). No matter how "beautiful" the respective interface, this GUI is now more and more replaced by a Conversational User Interface (CUI). Other still evolving interface styles are less text- and voice-driven, and therefore limit the messaging element to certain basic functions such as taking photos with the blink of an eye (smart glasses or smart cameras such as Blincam can do that today) but will eventually allow for richer interaction gestures (see project Soli). When coupled with an input-output feedback loop, so-called bionic lenses also hold a promising future.

ICU readmission is associated with longer hospitalization, mortality and adverse outcomes. An early recognition of ICU re-admission can help prevent patients from worse situation and lower treatment cost. As the abundance of Electronics Health Records (EHR), it is popular to design clinical decision tools with machine learning technique manipulating on healthcare large scale data. We designed data-driven predictive models to estimate the risk of ICU readmission. The discharge summary of each hospital admission was carefully represented by natural language processing techniques. Unified Medical Language System (UMLS) was further used to standardize inconsistency of discharge summaries. 5 machine learning classifiers were adopted to construct predictive models. The best configuration yielded a competitive AUC of 0.748. Our work suggests that natural language processing of discharge summaries is capable to send clinicians warning of unplanned 30-day readmission upon discharge.

Patient summarization is essential for clinicians to provide coordinated care and practice effective communication. Automated summarization has the potential to save time, standardize notes, aid clinical decision making, and reduce medical errors. Here we provide an upper bound on extractive summarization of discharge notes and develop an LSTM model to sequentially label topics of history of present illness notes. We achieve an F1 score of 0.876, which indicates that this model can be employed to create a dataset for evaluation of extractive summarization methods.

Amy works just like a human assistant, except she's not human. It's an AI bot made by X.ai, a company specializing in scheduling assistants that respond to natural language. Amy is so good at what she does that I find myself thanking her for booking a meeting, forgetting she needs no more thanks than my microwave. It's impossible to ignore all the buzz about AI bots. Last month, Facebook's David Marcus announced that over 30,000 bots have been built since the opening of its Messenger app to bot developers in April.

Recently, I needed to book a lunch meeting. To help coordinate, I asked Amy to assist and cc'd her on the email. "Amy," I wrote, "please help us find a time to meet. Let's plan for sushi at Tokyo Express on Spear Street." Amy looked at my calendar, found an open time suitable for everyone invited, and booked the meeting.