[developers] Extracting surface form of tokens from derivation trees

[Ann Copestake]

Thanks all!

now that 1212 is downloadable (thanks!) it seems that the first part
of the output provides the initial tokens and also gives their
cfrom:cto - hence it doesn't look hard to take the MRS and relate the
cfrom:cto in the MRS to those token numbers - the token feature
structure rep now seems unnecessary for what we want to do.  Or?  I am
not completely sure I understand the full space of possibilities as to
what could be involved in terms of the mapping, but it may be simplest
to just run it and find out,

Thanks,

Ann