The output of the tokeniser is a string of tokens, possibly with type labels (lexical property labels for use in parsing) attached.
For sentence parsing, the output of the analyser and parse constructor is, classically, a tree with labelled nodes, such as the trees familiar from phrase structure analysis. However other structures may be output, such as trees with labelled branches, nested feature structures, other kinds of graph, query language expressions for a database application, ...
For phonological parsing, it is doubtful whether trees are useful, or overly
complex, forcing artefactual decisions about, for instance, syllabification.
Note the unhappy `ambisyllabic' victims of phonological divorce, as in
.
Or is it 
?
Or 
?
Or 
?