Automatic article restoration

  • Authors:
  • John Lee

  • Affiliations:
  • Spoken Language Systems MIT Computer Science and Artificial Intelligence Laboratory, Cambridge, MA

  • Venue:
  • HLT-SRWS '04 Proceedings of the Student Research Workshop at HLT-NAACL 2004
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

One common mistake made by non-native speakers of English is to drop the articles a, an, or the. We apply the log-linear model to automatically restore missing articles based on features of the noun phrase. We first show that the model yields competitive results in article generation. Further, we describe methods to adjust the model with respect to the initial quality of the sentence. Our best results are 20.5% article error rate (insertions, deletions and substitutions) for sentences where 30% of the articles have been dropped, and 38.5% for those where 70% of the articles have been dropped.