Selected work

Representative Papers

Symphony for Speech-to-Text: Supporting Real-Time Medical Voice Interfaces

May 15, 2026

This work addresses the challenges of medical speech recognition, including domain-specific terminology, contextual ambiguity, and accurate transcription of clinical abbreviations and numerical values—issues that existing systems struggle to reconcile with real-time performance, accuracy, and generalization. The authors propose a modular decoupled architecture that separates the transcription pipeline into three specialized stages: domain-adapted recognition, formatting, and context-aware correction. This approach achieves, for the first time, high-recall recognition of medical terms and generates structured clinical text while supporting adaptive deployment across diverse scenarios. The system offers a production-grade API compatible with real-time dictation, conversational input, and batch processing. Evaluated on public medical speech datasets, it significantly outperforms state-of-the-art methods in the clinical domain while matching or exceeding their performance on general-domain tasks. The study also introduces the first Chinese clinical speech benchmark dataset to advance research in this area.

0 citationsRead paper

Symphony for Medical Coding: A Next-Generation Agentic System for Scalable and Explainable Medical Coding

Mar 31, 2026

This work addresses the longstanding reliance on manual medical coding, which is inefficient and error-prone, and overcomes key limitations of existing automated approaches—namely poor generalizability to new coding systems and lack of interpretability. The authors propose an agent-based reasoning framework that emulates human expert decision-making by dynamically retrieving official coding guidelines and integrating them with clinical text understanding, enabling adaptation to any coding system without retraining. This approach achieves zero-shot cross-system transfer—the first of its kind—and provides traceable justifications linking predicted codes to supporting evidence in the source documents. Evaluated across five real-world and publicly available datasets spanning multiple countries and clinical specialties, the method demonstrates state-of-the-art performance and strong practical deployability.

0 citationsRead paper

A medical coding language model trained on clinical narratives from a population-wide cohort of 1.8 million patients

Feb 27, 2026

This study addresses the inefficiency and systematic undercoding of secondary diagnoses in manual medical coding by developing a multimodal language model trained on 5.8 million electronic health records from 1.8 million patients in eastern Denmark—a population-scale cohort encompassing nearly all medical specialties. The model integrates clinical notes, medication records, and laboratory data to predict ICD-10 codes. Evaluated on a hold-out set of 270,000 patients, it achieves a micro-averaged F1 score of 71.8% and a top-10 recall of 95.5%. It also identified thousands of cases with missed secondary diagnoses, 76–86% of which were confirmed as valid upon manual review. The approach can automate approximately 50% of coding tasks, offering a scalable tool for epidemiological and multimorbidity research.

0 citationsRead paper

A Text-To-Text Alignment Algorithm for Better Evaluation of Modern Speech Recognition Systems

Sep 29, 2025

In conventional ASR evaluation, word error rate (WER) obscures critical errors involving rare words, named entities, and domain-specific terminology, hindering fine-grained error analysis. To address this, we propose a novel text alignment algorithm that integrates dynamic programming with beam-search scoring, significantly improving token-level matching fidelity—particularly for low-frequency and semantically sensitive units—between reference transcripts and ASR hypotheses. Unlike standard edit-distance-based alignment, our method enables high-fidelity, context-aware alignment in complex semantic scenarios, thereby exposing high-impact errors otherwise diluted in WER. Experiments demonstrate substantial gains in alignment accuracy over baseline methods, enabling reliable error attribution and model diagnostics. The algorithm is publicly available on PyPI, facilitating reproducibility and integration into existing evaluation pipelines.

0 citationsRead paper

Code Like Humans: A Multi-Agent Solution for Medical Coding

Sep 04, 2025

Medical coding requires mapping unstructured clinical text to over 70,000 ICD-10 diagnosis and procedure codes, with low accuracy and poor coverage for rare diseases representing a critical bottleneck. This paper proposes the first multi-agent generative coding framework designed for the full ICD-10 taxonomy: specialized agents emulate core expert decision steps—including terminology standardization, rule-based reasoning, and guideline verification—collaboratively adhering to official coding conventions. We introduce discriminative fine-tuning to enhance recognition of frequent codes and systematically identify model blind spots for rare codes. Experiments demonstrate state-of-the-art performance on rare diagnosis code prediction, the first end-to-end support for the complete ICD-10 code set, and the first empirical quantification of inherent systematic under-coding bias in generative models—revealing persistent gaps in rare-code recall despite high overall accuracy.

0 citationsRead paper

Recent publications

Latest Papers

Symphony for Speech-to-Text: Supporting Real-Time Medical Voice Interfaces

May 15, 2026

This work addresses the challenges of medical speech recognition, including domain-specific terminology, contextual ambiguity, and accurate transcription of clinical abbreviations and numerical values—issues that existing systems struggle to reconcile with real-time performance, accuracy, and generalization. The authors propose a modular decoupled architecture that separates the transcription pipeline into three specialized stages: domain-adapted recognition, formatting, and context-aware correction. This approach achieves, for the first time, high-recall recognition of medical terms and generates structured clinical text while supporting adaptive deployment across diverse scenarios. The system offers a production-grade API compatible with real-time dictation, conversational input, and batch processing. Evaluated on public medical speech datasets, it significantly outperforms state-of-the-art methods in the clinical domain while matching or exceeding their performance on general-domain tasks. The study also introduces the first Chinese clinical speech benchmark dataset to advance research in this area.

0 citationsRead paper

Symphony for Medical Coding: A Next-Generation Agentic System for Scalable and Explainable Medical Coding

Mar 31, 2026

This work addresses the longstanding reliance on manual medical coding, which is inefficient and error-prone, and overcomes key limitations of existing automated approaches—namely poor generalizability to new coding systems and lack of interpretability. The authors propose an agent-based reasoning framework that emulates human expert decision-making by dynamically retrieving official coding guidelines and integrating them with clinical text understanding, enabling adaptation to any coding system without retraining. This approach achieves zero-shot cross-system transfer—the first of its kind—and provides traceable justifications linking predicted codes to supporting evidence in the source documents. Evaluated across five real-world and publicly available datasets spanning multiple countries and clinical specialties, the method demonstrates state-of-the-art performance and strong practical deployability.

0 citationsRead paper

A medical coding language model trained on clinical narratives from a population-wide cohort of 1.8 million patients

Feb 27, 2026

This study addresses the inefficiency and systematic undercoding of secondary diagnoses in manual medical coding by developing a multimodal language model trained on 5.8 million electronic health records from 1.8 million patients in eastern Denmark—a population-scale cohort encompassing nearly all medical specialties. The model integrates clinical notes, medication records, and laboratory data to predict ICD-10 codes. Evaluated on a hold-out set of 270,000 patients, it achieves a micro-averaged F1 score of 71.8% and a top-10 recall of 95.5%. It also identified thousands of cases with missed secondary diagnoses, 76–86% of which were confirmed as valid upon manual review. The approach can automate approximately 50% of coding tasks, offering a scalable tool for epidemiological and multimorbidity research.

0 citationsRead paper

A Text-To-Text Alignment Algorithm for Better Evaluation of Modern Speech Recognition Systems

Sep 29, 2025

In conventional ASR evaluation, word error rate (WER) obscures critical errors involving rare words, named entities, and domain-specific terminology, hindering fine-grained error analysis. To address this, we propose a novel text alignment algorithm that integrates dynamic programming with beam-search scoring, significantly improving token-level matching fidelity—particularly for low-frequency and semantically sensitive units—between reference transcripts and ASR hypotheses. Unlike standard edit-distance-based alignment, our method enables high-fidelity, context-aware alignment in complex semantic scenarios, thereby exposing high-impact errors otherwise diluted in WER. Experiments demonstrate substantial gains in alignment accuracy over baseline methods, enabling reliable error attribution and model diagnostics. The algorithm is publicly available on PyPI, facilitating reproducibility and integration into existing evaluation pipelines.

0 citationsRead paper

Code Like Humans: A Multi-Agent Solution for Medical Coding

Sep 04, 2025

Medical coding requires mapping unstructured clinical text to over 70,000 ICD-10 diagnosis and procedure codes, with low accuracy and poor coverage for rare diseases representing a critical bottleneck. This paper proposes the first multi-agent generative coding framework designed for the full ICD-10 taxonomy: specialized agents emulate core expert decision steps—including terminology standardization, rule-based reasoning, and guideline verification—collaboratively adhering to official coding conventions. We introduce discriminative fine-tuning to enhance recognition of frequent codes and systematically identify model blind spots for rare codes. Experiments demonstrate state-of-the-art performance on rare diagnosis code prediction, the first end-to-end support for the complete ICD-10 code set, and the first empirical quantification of inherent systematic under-coding bias in generative models—revealing persistent gaps in rare-code recall despite high overall accuracy.

0 citationsRead paper