An Exploratory Study On The Appropriateness Of Latent Dirichlet Allocation For Automatic Discovery Of Product Associations From User-Generated Content
Johannes Putzke, Kai Fischbach, Detlef Schoder · European Conference on Information Systems · 2013
Latent Dirichlet Allocation (LDA) is a method that can be used to generate word association networks from unstructured text documents. However, no study has yet examined the applicability of LDA for deriving product associations from user-generated content. In this work, we apply LDA on 9,529 unstructured and uncategorized McDonald’s product reviews that were crawled from a German online review platform. We evaluate the applicability of LDA for deriving product associations from user-generated content. For this reason, we conducted a survey among 95 Information Systems undergraduate students about their associations with 17 McDonald’s-related nouns. Results indicate that LDA is a valid method for deriving product associations from user-generated content.