Prof. Dr. Walt Detmar Meurers
Publications
Data-Driven Correction of Function Words in Non-Native English

  

Adriane Boyd and Detmar Meurers

  

Proceedings of the 13th European Workshop on Natural Language Generation (ENLG) -- Helping Our Own (HOO) Challenge. Nancy, France. 267--269.

  

We extend the n-gram-based data-driven prediction approach (Elghafari, Meurers and Wunsch, 2010) to identify function word errors in non-native academic texts as part of the Helping Our Own (HOO) Shared Task. We focus on substitution errors for four categories: prepositions, determiners, conjunctions, and quantifiers. These error types make up 12% of the errors annotated in the HOO training data.

  

In our best submission in terms of the error detection score, we detected 67% of preposition and determiner substitution errors, 40% of conjunction substitution errors, and 33% of quantifier substitution errors. For approximately half of the errors detected, we were also able to provide an appropriate correction.

  


  

Electronically available:

  • Local copy: pdf (159.109 bytes)
  • ACL Anthology: pdf (166.483 bytes)

  

Note: The electronic versions of the publications linked on this page are the last versions I had the copyright for. Where a publisher copyedited and/or typeset the papers, the electronic copies linked here are NOT identical to the officially published version, which should be used for any quotes, references to page numbers, etc.

  


  

Bibtex entry:

 
@InProceedings{Boyd.Meurers-11,
  author    = {Boyd, Adriane and Meurers, Detmar},
  title     = {Data-Driven Correction of Function Words in Non-Native English},
  booktitle = {Proceedings of the 13th European Workshop
	      on Natural Language Generation -- Helping Our Own (HOO)
	      Challenge},
  year      = {2011},
  pages     = {267--269}
  address   = {Nancy, France},
  publisher = {Association for Computational Linguistics},
  url       = {http://purl.org/dm/papers/boyd-meurers-11.html}
}