What else can we do?
- Look at the data structure, currently we treat it as unstructured. What about weighting fields?
- Subset the expansion collections
- Find other expansion collections (SNOMED/UMLS?)
- Expand datasets with their repositories or associated publications