My research combines Natural Language Processing, domain-specific language resources, and evaluation. I am particularly interested in NLP for low-resource languages, especially Vietnamese, including systems that must handle specialized terminology, biomedical content, and code-switched speech.
Language data, multilingual translation, and text understanding.
I work on NLP problems where domain, language and data quality matter: Vietnamese–English machine translation, multilingual NLP, named-entity recognition, information extraction, corpus linguistics and NLP evaluation. I treat each project as one loop: define the task, build the data, evaluate the model, study its errors.
Language resources and models for specialized medical communication.
Biomedical NLP needs careful handling of terminology and annotation. My work covers medical machine translation (MedEV), biomedical NER with UMLS semantic types (En-ViMedNER) and medical code-switched speech (ViMedCSS).
LLM-based translation, prompt design, and evaluated workflows.
My LLM work covers LLM-based translation, prompt design and the evaluation of generated responses. I also build AI agent apps for planning, using AI coding agents under my specs and review. In both, the agent proposes and a person approves (human-in-the-loop).
Dataset construction, annotation quality, benchmarks, and evaluation.
I build language datasets as part of the research: corpus construction, annotation design, inter-annotator agreement, quality checks, benchmark design and model evaluation. Examples: En-ViMedNER, ViMedCSS, MedEV, the ViLEC learner corpus and the 100M-word VnC 1.0.
