All research areas
Research
Natural Language Processing
Language systems that work outside English and outside the demo.
We work on text where the easy assumptions break: low-resource languages, code-switching, domain jargon, and OCR noise. Bengali NLP in particular is under-served, and it is close to home for us.
In production
CMS — reviewer matching & abstract review
Our NLP work powers CMS's topic-based reviewer matching and conflict-of-interest detection — the clearest line from a paper here to a feature in production.
See the product →What this looks like in practice
- Information extraction and document understanding
- Classification, summarisation, and semantic search
- Low-resource and Bengali language modelling
- Named entity recognition for specialised domains
- Corpus construction and annotation pipelines