arXiv subjects

Stephen Ma

Publications and source records attributed to Stephen Ma.

4 recordsLinked to original sources

arXiv2026

Multi-Method Validation of Large Language Model Medical Translation Across High- and Low-Resource Languages

Language barriers affect 27.3 million U.S. residents with non-English language preference, yet professional medical translation remains costly and often unavailable. We evaluated four frontier large language models (GPT-5.1, Claude Opus 4.5, Gemini 3 Pro, Kimi K2) translating 22 medical documents into 8 languages spanning high-resource (Spanish, Chinese, Russian, Vietnamese), medium-resource (Korean, Arabic), and low-resource (Tagalog, Haitian Creole) categories using a five-layer validation framework. Across 704 translation pairs, all models achieved high semantic preservation (LaBSE greater than 0.92), with no significant difference between high- and low-resource languages (p = 0.066). Cross-model back-translation confirmed results were not driven by same-model circularity (delta = -0.0009). Inter-model concordance across four independently trained models was high (LaBSE: 0.946), and lexical borrowing analysis showed no correlation between English term retention and fidelity scores in low-resource languages (rho = +0.018, p = 0.82). These converging results suggest frontier LLMs preserve medical meaning across resource levels, with implications for language access in healthcare.

cs.CL↗

arXiv2025

SHACL-SKOS Based Knowledge Representation of Material Safety Data Sheet (SDS) for the Pharmaceutical Industry

We report the development of a knowledge representation and reasoning (KRR) system built on hybrid SHACL-SKOS ontologies for globally harmonized system (GHS) material Safety Data Sheets (SDS) to enhance chemical safety communication and regulatory compliance. SDS are comprehensive documents containing safety and handling information for chemical substances. Thus, they are an essential part of workplace safety and risk management. However, the vast number of Safety Data Sheets from multiple organizations, manufacturers, and suppliers that produce and distribute chemicals makes it challenging to centralize and access SDS documents through a single repository. To accomplish the underlying issues of data exchange related to chemical shipping and handling, we construct SDS related controlled vocabulary and conditions validated by SHACL, and knowledge systems of similar domains linked via SKOS. The resulting hybrid ontologies aim to provide standardized yet adaptable representations of SDS information, facilitating better data sharing, retrieval, and integration across various platforms. This paper outlines our SHACL-SKOS system architectural design and showcases our implementation for an industrial application streamlining the generation of a composite shipping cover sheet.

cs.AI↗

arXiv2024

Fine Tuning Large Language Models for Medicine: The Role and Importance of Direct Preference Optimization

Large Language Model (LLM) fine tuning is underutilized in the field of medicine. Two of the most common methods of fine tuning are Supervised Fine Tuning (SFT) and Direct Preference Optimization (DPO), but there is little guidance informing users when to use either technique. In this investigation, we compare the performance of SFT and DPO for five common natural language tasks in medicine: Classification with text data, Classification with numeric data, Clinical Reasoning, Summarization, and Clinical Triage. We find that SFT alone is sufficient for Classification with text data, whereas DPO improves performance for the more complex tasks of Clinical Reasoning, Summarization and Clinical Triage. Our results establish the role and importance of DPO fine tuning within medicine, and consequently call attention to current software gaps that prevent widespread deployment of this technique.

cs.CL↗

arXiv2012

Competing Process Hazard Function Models for Player Ratings in Ice Hockey

Evaluating the overall ability of players in the National Hockey League (NHL) is a difficult task. Existing methods such as the famous "plus/minus" statistic have many shortcomings. Standard linear regression methods work well when player substitutions are relatively uncommon and scoring events are relatively common, such as in basketball, but as neither of these conditions exists for hockey, we use an approach that embraces the unique characteristics of the sport. We model the scoring rate for each team as its own semi-Markov process, with hazard functions for each process that depend on the players on the ice. This method yields offensive and defensive player ability ratings which take into account quality of teammates and opponents, the game situation, and other desired factors, that themselves have a meaningful interpretation in terms of game outcomes. Additionally, since the number of parameters in this model can be quite large, we make use of two different shrinkage methods depending on the question of interest: full Bayesian hierarchical models that partially pool parameters according to player position, and penalized maximum likelihood estimation to select a smaller number of parameters that stand out as being substantially different from average. We apply the model to all five-on-five (full-strength) situations for games in five NHL seasons.

stat.AP↗