CSCAR

This workshop introduces the possibilities for using quantitative methods to study documents that are typically treated only qualitatively, such as novels, newspaper articles, judicial opinions, and web pages.  These statistical methods will facilitate the formulation and exploration of research questions arising from studies of digitized but untagged texts. No prerequisites are required. No previous experiences with text analyses and/or statistical analyses are assumed.

Research and documents from business, law, policy, indeed, all humanistic and social scientific fields rely ever more on quantitative analyses of digitized texts, be they historical archives, legal documents, customer surveys, or poetry collections online. Often the amount of information is enormous, daunting analysts to review by eye or code by hand. Computer-aided methods of processing and analyzing textual data solve this problem and allow previously impossible discoveries. But the methods are unfamiliar to most people who could make good use of text mining. This workshop introduces concepts and teaches skills for mining and analyzing texts. It begins with key statistical concepts and graphical tools that are commonly used in analyzing textual data, such as word frequencies, standardized token/type ratio, and relationship statistics on word co-occurrences. The workshop also covers text processing skills, for example, preparing texts for analysis, developing stop-word lists and lemma (head-word) lists, and porting results for further processing in other programs. We will explore WordSmith (http://www.lexically.net/wordsmith/), a program using a capacious array of text study techniques, for example, producing interactive concordances; constructing concept sets for document and corpus analyses; applying statistics to understand definitions, styles, and thematic concerns both at one point and over time. Participants will have a chance to practice these skills using digitized texts provided by the workshop. The workshop will explore and entertain new questions about the Bible, the United States Constitution and its Amendments, and 2008 Democratic and Republican Convention speeches. An important local example, using radiology reports, will illustrate how text mining can be used as an expert system.