SRCLD Presentation Details
  Title  
       
    Accuracy of computerized grammatical category coding  
Author(s)
Ron Channell - Brigham Young University
Karen Johnson - Brigham Young University

SRCLD Info
SRCLD Year: 2010
Presentation Type: Poster Presentation
Poster Number:
Presentation Time: (na)
Categories
- Assessment
Abstract
     A computer program for automated grammatical category ("part of speech") coding of SALT-formatted child language sample transcripts might be useful if accurate, easy to use, and inexpensive. A program called GCS is available as a freeware download; this poster examines its accuracy.
     All words in 50 language sample transcripts from typically developing children (ages 2;6 to 11;0) were manually coded as to LARSP grammatical category. Samples were then coded by the GCS software, and these two codings were compared on a word-by-word basis.
     The overall accuracy was 97%. GCS was at least 99% accurate in coding the categories initiator, determiner, noun, pronoun-personal, verb, auxiliary-modal, negation, copula, and preposition. It was at least 95% accurate in coding the categories adjective, pronoun-other, auxiliary-other, adverb, existential there, and subordinator. It was 90% accurate in coding wh-question words (confusing them with subordinators 9% of the time), 88% on particles (confused with prepositions 12%), and 78% in coding intensifiers (confused with adverbs 9%, pronoun-other 5%, and several other categories).