Automated developmental sentence scoring using computerized profiling software

This study examined the accuracy of fully automated Developmental Sentence Scoring (DSS; L. L. Lee, 1974) analysis performed by the Computerized Profiling (CP) software (S. H. Long, M. E. Fey, and R. W. Channell, 2000). Samples from 48 school-age children (28 with language impairment) yielded 9,084 utterances that were DSS coded both manually and by CP. The point-by-point agreement of CP with manual coding was 78%, with per-category levels of agreement ranging from 0% to 98%. Agreement levels were about 2% lower on samples from children with language impairment. Though significantly higher, the scores computed by CP were highly correlated (r =.97) with the manually computed scores. Further work on improving the accuracy of automated DSS analysis is warranted.