Measure of media coverage of a candidate for office

A researcher wanted to make a measure of media coverage of a candidate for office, and therefore created a set of coding rules to code words in newspaper articles as either pro” or “con” toward the candidate.

Instead of hiring students to implement these rules, however, the researcher used a computer to code the text, by counting the frequency with which certain words, were mentioned in a series of articles. What would be the reliability of such a computer-driven measurement strategy, and why?