“…This series of steps entailed splitting contractions into their constituent words, removing punctuation, removing numbers, converting text to lower case, removing excess spaces (in circumstances where a single space would be sufficient), and removing highly common terms (called "stop words") that would not provide considerable benefit for differentiating between the topics of interest (e.g., "for," "a," "the," "it," "that," and "to"). Details for these procedures have been described elsewhere (e.g., Lin & He, 2009;Zhao et al, 2011) and are important steps to produce interpretable topic modeling results. Following preprocessing, we created a documentterm matrix (DTM) representing the frequency of terms across each of the documents in the corpus.…”