A Collaborative Resource for Practical, Shareable Clinical NLP Tools
Creating a shared place to:
- consolidate oncology-specific extraction targets
- identify simple, robust solutions
- map community expertise
- share ready-to-use notebooks, heuristics, regexes, and workflows
- avoid duplicative effort across sites
- How do we best curate all the mapping targets in a single location for maintainability?
- As a start, this is mostly a collection of pointers, notes, conventions
- need a straightforward way to provide minimal working examples to support their use
- For logical units of work it would be better to apply in-code documentation
- at this point the structure is quite loose and therefore the most sensible location for this isn't clear
- strategic decision-making required
- medspacy customisations for tokenisation / sectionising are sufficiently nuanced to not be just
config files- create registerable spacy language with these customisations - where will it live?
- value-extractor + label-extractor classes as PR?