[developers] Roots in ACE output

Stephan Oepen oe at ifi.uio.no
Tue Aug 19 19:20:03 CEST 2014


colleagues,

> [...] an interpretation of the set of root symbols as an ordered list, as another way to `resolve' the ambiguity where more than one root symbol would admit an analysis.

i would even be prepared to claim that interpretation as a property of
our current formalism :-).  often (subsets of) the root symbols are
ordered by specificity, and on this view a relevant property of each
complete analysis is the top-ranked (most specific) root symbol that
is compatible with its TFS.

i do agree with woodley, however, that the formal status of the root
symbol when recorded in the derivation tree is slightly different from
nodes recording grammar rules or lexical entries.  in the [incr
tsdb()] Unified Derivation Format (UDF), root nodes are syntactically
distinguished from other nodes, and are considered optional.

on the other hand, in assessing the consistency of a derivation, all
of the pieces need to unify (have their constraints satisfied),
including the root symbol.  we have at times used the term ‘grammar
entity’ to refer to the union of lexical entries, grammar rules, and
root symbols.  thus, to me it still seems defensible to record the
property i allude to in the first paragraph directly as part of the
derivation tree.

best wishes, oe



More information about the developers mailing list