Case Study: Understanding Natural Language - Thinking as Computation - page 167

Figure 8.10. Connecting a partial trace to a parse tree

np2([man,in,the,park,with,a,tree],_)
common_noun(man,_),
mods([in,the,park,with,a,tree],_)
mods([in,the,park,with,a,tree],_)
pp([in,the,park,with,a,tree],_),
mods([],_)
preposition(in,_,_),
np([the,park,with,a,tree],_)
np([the,park,with,a,tree],_)
article(the), np2([park,with,a,tree],_)
np2([park,with,a,tree],_)
common_noun(park, ), mods([with,a,tree], )
mods([with,a,tree],_)
pp([with,a,tree],_), mods([],_)
preposition(with,_,_), np([a,tree],_)
np([a,tree],_)
article(a), np2([tree],_)
np2([tree],_)
common_noun(tree,_), mods([],_)

NP2

noun Mods

PP Mods

prep NP

article NP2

noun Mods

PP Mods

prep NP

article NP2

noun Mods

So the thinking here is not only about the words and the grammar, but also about
what the words mean.

Reconstructing parse trees

∗

There is a close connection between these traces and parse trees. It is possible with
some effort to reconstruct a parse tree from the trace of a successful query to the
parser. Look at figure 8.10. On the left-hand side of the figure are the Call parts of a
successful trace of an np2 ; on the right, is the corresponding parse tree.

Some additional queries

Some additional queries and their answers are shown in figure 8.11. Note that noun
phrases can be used to identify hats as easily as people, as the second query shows.