Jason B. Jones (Trinity College) wrote a ProfHacker post discussing the JSTOR Labs Text Analyzer, a new tool that allows researchers to locate relevant related articles from the JSTOR archive by uploading a source text of the user’s choice.
JSTOR says that it compares the text to a list of 40,000 topics and a set of human-curated rules. As I look at this list of terms, I worry a little that none of the words I’d say are most important (interpretation, deferred action, time, temporal, etc). Meanwhile, words that get used *once*, in clearly attributive ways, such as “castration,” are identified as key topics. Likewise, the “Rotary Club” is posited as a relevant organization, which is a little grandiose for the joke I made.
The list of articles is decently relevant, but probably not enough for me to want to immediately download them as spot-on to my interests. However, this is dependent on the “prioritized terms,” so if I selected different ones this would improve.
dh+lib readers who manage electronic resources in GLAM environments will appreciate the way JSTOR’s approach serves to surface subscribed resources through digital humanities-inflected methods and represents an approach to locating sources that is less keyword-based.