SYNTACTIC Features

SYNTACTIC Features

1. features tend to produce noisy data set result from unavoidable error by parser

2. require robust and accurate NLP tools i.e. parser or POS tagger to perform analysis

3. syntactic information are more reliable authorial fingerprint

4. authors tend to use syntactic features unconciously

5. Rewrite rule frequencies

   • so called syntactic dependency frequencies
   • perform better than vocabulary richness and lexical measures
   • produce by full parse tree of each sentence
   • describe syntactic class of each word and how the word combine to form phrases etc.
   • E.g.: A:PP -> P:PREP + PC:NP

6. Sentence Chunks

   • measures
     • noun phrase counts
     • verb phrase counts
     • length of noun phrase
     • length of verb phrase
   • extracted automatically by NLP tools with relatively high accuracy
   • simpler than rewrite rules
   • detect sentences and phrases boundaries
   • E.g.: NP[Another attempt] VP[to exploit] NP[syntactic information] VP[was proposed] PP[by E.Stamatatos]

7. Sentence and Phrase structure

   • require tools to perform partial parsing
   • Partial parsing provide syntactic features of varying complexity
   • output from partial parser are transformed into an ordered stream of syntactic labels
   • E.g.: NX DT JJ NN (In a Noun Phrase consist of a Determiner, a Adjective and a Noun)

8. POS Tag

   • provide only the hint of the structural analysis of sentences
     • how the words are combined to form a phrase?
     • how are the phrases are combined into higher level structures
   • require POS tagger as tool
   • perform quite accurate in unrestricted text
   • morpho-syntactic tagger + syntactic analyzer enable extraction of uni,bi and trigram of POS tags
   • measures
     • POS tag frequencies
     • POS tag n-grams frequencies

9. Syntactic errors

   • run-on sentences
   • mismatched Tense
   • ortographic errors
   • sentence fragments
   • spell checker accuracy issue