The challenge of measuring ideological bias in written digital media

This paper makes a proposal to measure the ideological bias
of digital media that is based on machine learning. We use
a strategy based on the use of texts to identify ideologically
charged words, which studies of political science also use to
measure the positions of parties and candidates. Our proposal
presents two differential features with respect to previous
studies: it uses the concept of a frame as unit of analysis to
identify ideological bias and it relies on the tweets of politicians
as the reference text for identifying ideologically connected
groups of word – i.e., frames.