The idea

• Aim: Detect a wide range of syntax differences

• Tag 2 or more collections of comparable material (using an automatic POS-tagger):

  • we did: learners, or 2nd language speakers versus natives
  • many other options: different regions or discourses

• Take n-grams (3-5grams) of POS-tags

• Statistically compare their frequencies

• Sort the significant POS-n-grams by weight (frequency or extent of difference)

Notes: