ABSTRACT: We previously developed an automated quality rating technique (AQA) for depression websites and showed that it correlated 0.85 with expert ratings using evidence-based guidelines. AQA scores the pages within a site against complex learned 'quality' and 'relevance' queries using the BM25 ranking function. It then aggregates these scores across sites using a learned combining function. In this talk, we report on our generalisation of the AQA method and our evaluation of its application to a different health domain, namely obesity. We find that correlations as high as 0.80 averaged across ten folds can be obtained, but that performance is quite dependent on the choice of the high quality websites used for generating the 'quality' query. This is ongoing work ....
From: Automated Quality Assessment of Obesity Websites, School of Computer Sciecne, ANU, 2009
Post a Comment