Paper: Using Latent Dirichlet Allocation for Child Narrative Analysis

ACL ID W13-1914
Title Using Latent Dirichlet Allocation for Child Narrative Analysis
Venue Workshop on Biomedical Natural Language Processing
Session
Year 2013
Authors

Child language narratives are used for lan- guage analysis, measurement of language development, and the detection of lan- guage impairment. In this paper, we ex- plore the use of Latent Dirichlet Alloca- tion (LDA) for detecting topics from nar- ratives, and use the topics derived from LDA in two classification tasks: automatic prediction of coherence and language im- pairment. Our experiments show LDA is useful for detecting the topics that corre- spond to the narrative structure. We also observed improved performance for the automatic prediction of coherence and lan- guage impairment when we use features derived from the topic words provided by LDA.