Knowing your data: analysis and visualization as data exploration

Data analysis and visualization does not exclusively come after putting together corpora or text datasets, defining research questions and committing to methods of addressing them. Sometimes it is really important to use analysis and visualization tools as a mean of better understanding our data right from the start. Visualizing part of a dataset, running a query, testing a hypothesis, may reveal interesting patterns we hadn't noticed before, help us highlight entities and relations, or even expose problems with the data we were not aware of. A graph of a text-based dataset, for example, showing frequencies of terms in a document is useful not just in terms of testing research questions and hypotheses about term frequencies but also because it can reveal the conditions (of authorship, control, omission) and methodologies in which the dataset has been produced in the first place.