Project members:

Birgitta Sahlén, Department of Logopedics, Phoniatrics & Audiology, Lund University Sverker Sikström, Department of Psychology, Lund University Rasmus Ahrlinger Bååth, Cognitive Science, Lund University Simone Löhndorf, Centre for Languages and Literature, Lund University, Kristina Hansson, Department of Logopedics, Phoniatrics & Audiology, Lund University, Viktoria Åkerlund, Department of Logopedics, Phoniatrics & Audiology, Lund University

The purpose of this project is to apply a semantic space method, Latent Semantic Analysis (LSA), to quantify the semantic content in narratives from children with typical and impaired language development in order to evaluate if it can be used as a tool to assess the development of semantic representations and if it can be used to identify children with language impairment. Most studies of typical and impaired language development focus on form (phonology, grammar) and in studies on content there is a bias towards studying lexical skills on a word-by-word basis in experimental studies. Arguably the inclusion of multidimensional semantic features may yield a more accurate description of linguistic maturity in data on the text level. So far, no assessment instrument exists that can capture this aspect of language skills. In a first step we have devised a model to quantify semantic linguistic maturity (SELMA) based on a high dimensional semantic representation of words created from the co-occurrence of words in a large text corpus. A comparison between the SELMA measure of semantic maturity (i.e., age predicted by LSA), with maturity ratings by speech and language pathology students showed high agreement between SELMA and the human raters, higher than between SELMA and chronological age. We can thus show that SELMA is a valid measure of semantic maturity. In a next step we use LSA to predict whether a child has language impairment or not. Based on narratives from 80 children with language impairment and 80 age-matched controls, the preliminary findings are that LSA is significantly more successful than chance in identifying which children belong to the language impaired group and which children belong to the control group.

Publications:

Hansson, K., Bååth, R., Löhndorf, S., Sahlén, B. & Sikström, S. (revised and resubmitted manuscript). Quantifying Semantic Linguistic Maturity (SELMA) in children.

Bååth, R., Sikström, S., Kalnak, N., Hansson, K. & Sahlén, B. Using latent semantic analysis to identify children with specific language impairment. (Manuscript to be submitted).

Magnusson, V. (2011). Validering av SELMA (Semantic Linguistic Maturity) i barns muntliga berättelser. Magisterarbete i Logopedi, Lunds Universitet.