Data collected by Day of DH moderators in response to “How do you define digital humanities?” is provided for free at\u00a0whatisdigitalhumanities.com. This data set was downloaded in .csv format. First, the entire dataset was run through Voyant, the online text analysis tool, in order to analyze word frequency for the entire corpus. Due to the fact that later analysis would be run on each year individually, and each year\u2019s document was drastically different in length, all measurements of word frequency were taken using relative frequency (i.e. the frequency of occurrence out of a scaled-up corpus of 1M words) as opposed to raw count (i.e. the actual frequency of occurrence), and rounded to the nearest whole numbers. Common English language stop words were used; the terms \u201cdigital,\u201d \u201chumanities,\u201d and \u201cdh\u201d were also removed from the dataset, as they were outliers in terms of frequency and also deemed to be unrevealing in terms of textual insight, given that they were most often used in the context of referring to that which respondents were asked to define. Generic works such as \u201cuse,\u201d \u201cwork,\u201d and \u201cway\u201d were also removed. In order to be more encompassing and to more accurately reflect the use of these words in different forms across the document, terms were converted to strings as applicable. For instance, technology <\/em>and technologies <\/em>were merged into the string technolog*<\/em>, which then included such terms as technological <\/em>and technologically. <\/em>Similarly, research<\/em> became research<\/em>* to include researching<\/em> and researchers<\/em>; methods<\/em> became method*<\/em> to include methodological<\/em> and methods<\/em>; computing<\/em> became comput<\/em>* to include computers<\/em>, computational<\/em>, etc.<\/p>\n The top twenty words were then coded as describing either M<\/em>ethod<\/em>,\u00a0Practice<\/em>, or V<\/em>alue<\/em>. The categorizations are outlined in Table 1 in the Findings section, below.<\/p>\n Given that the dataset specified the year of collection for each response, data were then separated into individual documents by year in order to run individual analysis on each time period. Each year\u2019s collection of definitions was analyzed, again using Voyant, using the same stop words as with the main dataset. Likewise, words in different formats and tenses were combined using appropriate strings. Word frequencies were taken for each word in each year and graphed to show the evolution of that term across the entire corpus.<\/p>\n Findings and Discussion<\/strong><\/p>\n Table 1: Category, Term, and Relative Frequencies of Top 20 Terms in Corpus<\/p>\n
<\/a> <\/div>\n