Feature Analysis for Aphasic or Abnormal Language Caused by Injury

The interest of researching alterations in aphasic or abnormal language has maintained and keeps growing, following a diversity of approaches. Particularly, the period corresponding to the post-traumatic recovery stage in impaired language caused by a traumatic brain injury has been scarcely explored. The findings reported here specify the steps followed during the inspection of a lexical feature set with the objective of determining which contribute most to describe TBI language singularities throughout the first two stages of recovery (after three and six months). For this purpose, we employ language technologies, statistical analysis, and machine learning. Starting with a 25-feature set, after conducting an analytical process, we attained three selected sets with fewer attributes, having comparable efficacy measures to those obtained from the complete feature set taken as baseline.