Exploring Political Ideologies of Senators With Semantic Analysis Tools: Further Validation of CASS

Phrase counting is an effective approach to capturing individual differences in language use. Specific phrases (e.g., “war on terrorism”) powerfully predict a congressperson’s political ideology. The question addressed in this study is whether there is additional information that can be extracted from the indirect relations among words in large Senatorial speech databases. Given that direct co-occurrence of target words is a very low likelihood event, we focus on higher-order co-occurrence (e.g., whether two target words appear in similar semantic contexts), using the free software Contrast Analysis of Semantic Similarity (CASS) to compute individual differences (www.casstools.org). We describe how we used CASS in detail and provide a tutorial. Using text transcripts from 86 Senators, totaling over 150 million words, we demonstrate that CASS can account for political ideology above and beyond phrase counts. By complementing phrase count methods, CASS may be a useful method for the digital humanities and social sciences more generally.