This kind of phonological ambiguity is
frequent enough, also because sound may be distorted. Actually, it has
been considered important enough in speech processing that a specific
representational device was created for it. It is the so called "word
lattice" used to represent in a compressed way the various sequences
of words that could correspond to the phonological sequence, each word
corresponding to a subpart of that sequence. Note that the different
sequences of word may imply different way of cutting of the string into parts
corresponding to words as for example (taken from wikipedia) "a
cruise, eh, lass?" and "accrues, hélas!".
Some parsing techniques can directly accomodate word lattices, as they
would a word sequence, and either eliminate some variants of the
phonological ambiguity because they cannot be made syntactically
correct, or treat them as a syntactic ambiguity problem, using
whatever means at their disposal for this purpose.
Exercise in phonological ambiguity have been pushed to an extreme with
holorime verses. A well known one in French is
Gall, amant de la Reine, alla, tour magnanime!
Galamment de l'Arène à la Tour Magne, à Nîmes.
You can fin several more at http://fr.wikipedia.org/wiki/Vers_holorimes.