I agree that if the program can recognise the patterns which indicate one of these parallel constructions (e.g. ‘respectively’), it would be easy to write a piece of code which matches them up. The curious thing I found here is just that it can’t be done from within the context-free grammar; it has got to happen in a second post-processing step based on the syntax tree and additional rules.

The parsing library I am using here is pyparsing (http://pyparsing.wikispaces.com/), a remarkably elegant library, but most libraries use something which is more or less context-free.

It looks like the first is different to the second in parsing terms. The first is recognisable as two sets of collerated data simply because in English ‘respectively’ is rarely used except to flag the construct. You could then fairly easily recognise the separators of ‘,’ and ‘and’.

The second is possibly easier to parse once you understand that it’s two sets of collerated data, but getting there could be tricky - and way too often people assume the reader can guess the measurement units from the context.

What are these typical parsing libraries?