Creole & lower-resourced NLP · Multilingual evaluation · Responsible security
Heather Lent
Assistant Professor
In my work on Creole and other lower-resourced languages, I take community needs—not simply inherited English benchmarks—as a starting point. Across my research, I study machine translation, transfer, multilingual data quality, language-model security, and the ethics of security research.
Heather’s research asks how language technology can serve languages and communities overlooked by mainstream NLP. She develops methods, datasets, benchmarks, and machine-translation systems for Creole and other lower-resourced languages, combining work on transfer, sociolinguistic variation, semantic parsing, and data quality with direct attention to what communities actually need.
She has also established a complementary strand in multilingual language-model security and research ethics. This work examines multilingual vulnerabilities and defences, alongside broader questions of harm minimisation and responsible disclosure. Across both strands, community considerations are part of the technical design—not an assessment added after a system is built.
Creole and lower-resourced NLP
Language modelling, transfer, machine translation, semantic parsing, and evaluation beyond the best-resourced languages.
Community-grounded evaluation
Letting community priorities and sociolinguistic realities change the tasks, datasets, benchmarks, and measures of success.
Multilingual security and ethics
Testing whether vulnerabilities, defences, and disclosure practices protect lower-resourced languages on their own terms.
Selected programmes
Projects and labs
Selected work
5 representative papers
How Good is Your Wikipedia? Auditing Data Quality for Low-resource and Multilingual NLP
Audits all non-English Wikipedia editions and shows why data quality—not language coverage alone—must be a first-class concern in multilingual NLP.
- Multilingual & Lower-Resource NLP
- Factuality, Reliability & Evaluation
NLP Security and Ethics, in the Wild
Audits ethical practice in NLP security and develops concrete guidance for harm minimisation and responsible disclosure.
- Security, Privacy & Safety
- Communities, Domains & Consequences
Text Embedding Inversion Security for Multilingual Language Models
- Security, Privacy & Safety
- Multilingual & Lower-Resource NLP
CreoleVal: Multilingual Multitask Benchmarks for Creoles
- Multilingual & Lower-Resource NLP
- Communities, Domains & Consequences
What a Creole Wants, What a Creole Needs
- Multilingual & Lower-Resource NLP
- Communities, Domains & Consequences
Supervision and collaboration
Questions worth working on
I am interested in projects on lower-resourced and multilingual NLP, Creole languages, machine translation, data and evaluation quality, multilingual security, and responsible research practice.