Seeing the Forest: Applying Latent Semantic Analysis to Smartphone Discourse

We apply latent semantic analysis (LSA) to understand how media discourse and cognition about smartphones evolved over time. LSA is a useful method to take advantage of large amounts of available text, discern meaning within the text, and see how meanings change over time, across the media coverage in the sector. We explain the theory of LSA, the process, and apply it to a dataset of over 83,000 media articles to create a semantic model of document and word meanings. We measure how groups of documents differ and then visualize how the discourse changes over time. We find that LSA is useful for measuring how discourse shifts across this broad set of data. In our empirical case, we find that smartphone discourse went through four distinct periods, with different dynamics of transition and stability. These characteristics suggested particular theoretical bases which LSA is also well suited to examine.