Words and networks: How reliable are network data constructed from text data?

Social network data as well as the information produced or shared by network participants are prominent sources for studying reputation and authority in social media. Research studies on this topic often start with one or more network datasets and bring relevant substantive questions about socio-technical concepts such as the evolution of credibility to the data. This chapter deals with the reliability of network data itself and aims to shed some light on the following question: How reliable or accurate are network data depending on the data construction method for cases where text data are used as an input to this process? I provide a concise overview on some of the most common methods for constructing network data from text data sources, report on our findings from applying these methods to three corpora from different domains and genres, and derive implications and suggestions for theoretical and practical work.