Wish list

Wish list for 'checklist diff' tool

Please read README first

• Documentation of method
• Try out more taxonomic sources e.g. CoL, MSW 2/3/4
• Translation to RCC-5
• Method
  • Handle 'unclassified' ('incertae sedis') containers and child rank inconsistencies
  • Compute RCC-5 > vs. the usual = when B-TNU has no exact A-TNU MRCA match
  • Eliminate NCBI 'containers' without losing implicit disjointness claims (e.g. via ranks or via 'mutexes')
  • Quantified uncertainty (confidence in RCC-5 judgments)...
  • Match even when the genus differs (if closely related, e.g. a sibling also matches in the same way, and unambiguous)
  • Match even when there is a gender error (-us vs. -a)
  • Enlist genetic and occurrence data, and perhaps other data, in understanding TNUs and/or assigning data records to TNUs
• Display
  • Fix display of synonyms; maybe 3 columns: accepted TNU in A hierarchy, a name that's in both A and B that enables the match (it's either A or an A synonym, and also either B or a B synonym), accepted TNU in B hierarchy
  • Stop eliding repeated names with =; maybe put name comparison status (= vs. blank) in a separate column
  • Show B nesting level somehow? Show A nesting better (indentation??)? Deal with ranks better? (NCBI ranks are not always consistent, sometimes not given at all)
  • Group A-TNU children according to their B-TNU parent, when the implied A-group is split into multiple B-groups
  • Maybe show all consistent RCC-5 options, as opposed to just suggesting one? Hmm...
  • Nice HTML output similar to Avibase checklist comparator
  • Graphviz? (that might be better done by separate fdownstream tools)
• Questions
  • Is there a sensible way to distinguish 'wholly new' from, say, splitting? Or to suggest to the user that either is a possibility for 'no match'? (in the case of Genbank, we could look at the sequence records maybe, as a heuristic??)