Determining syntactic complexity using very shallow parsing
Voss, Matthew John
MetadataShow full item record
This thesis describes a rule-based computer program, the Shallow Syntactic Complexity Analyzer (ShaC), for determining the syntactic complexity of English-language text. Syn-tactic complexity is determined by comparing strings of text to templates. The templates were constructed following the modi ed D-Level scale (Covington et al., 2004), which ranks syntactic complexity based on the age at which young children rst acquire various syntactic structures. The later a structure is acquired, the higher ranking it gets. ShaC is unique in that it attempts to give a good estimate of syntactic complexity without doing a deep syntactic analysis. Such a detailed analysis would be time consuming; ShaC uses heuristics and generalizations to greatly simplify the task at hand. The result is a quick and e cient method for estimating syntactic complexity. ShaC score correlates highly with D-Level. Sentences with a verb taking a nite complement, verbs taking an -ing complement, verbs taking a non- nite complement, and comparatives drive the correlation.