A0357
Title: Measuring technology acceptance over time by online customer reviews based transfer learning
Authors: Daniel Baier - University of Bayreuth (Germany) [presenting]
Andreas Karasenko - University of Bayreuth (Germany)
Alexandra Rese - University of Bayreuth (Germany)
Abstract: Online customer reviews (OCRs) are user-generated semi-formal evaluations of objects (brands, companies, products, services, technologies). They typically consist of a time stamp, a star rating (1 to 5 stars) of the evaluated object and, in many cases, a natural language comment that details the perceived strengths and weaknesses. Up to now, many methodological approaches have been developed and applied to analyze and aggregate OCRs, as well as to improve products and services based on this knowledge. So, a prior study applied a lexicographic text mining approach similar to sentiment analysis to OCRs of IKEAs augmented reality app. They predicted construct scores for the extended technology acceptance model (TAM) and validated these predictions by conducting an additional extended TAM survey among app users. A new transformers-based approach is presented for the same purpose. A transfer learning model is trained, tested, and validated based on large samples of OCRs and corresponding extended TAM construct scores given by experts. The results are promising. They go beyond conducting an extended TAM survey for an object by validly predicting the development of construct scores over time.
