Scale Development Using Twitter Data: Applying Contemporary Natural Language Processing Methods in IS Research

Abstract

The availability of big data sources and developments in computational linguistics present an opportunity for IS researchers to pursue new areas of inquiry and to tackle existing challenges with new methods. In this paper, a novel way of developing measurement scales using big data (i.e., tweets) and associated methods (i.e., natural language processing) is proposed and tested. The development of a new scale, the technology hassles and delights scale (THDS), is used to demonstrate how a syntax aware filtering process can identify relevant information from a large corpus of tweets to improve the content validity of a scale. In comparing themes generated from analyzing 146 million tweets, with themes generated from semi-structured interviews, a reasonable overlap is observed. Further, the potential for identifying even more relevant themes from within subsets of the tweet dataset is uncovered.