J. Cohen’s research while affiliated with Hillsdale College and other places

... The first approach entails using established cutoff points, which can be in an unstandardized metric (raw variable units) or in the familiar standardized metric. For a standardized example, a value for Cohen's d of 0.2 is commonly considered to be a small effect size, whereas d = 0.5 and d = 0.8 are considered medium and large, respectively (Cohen, 1988; see Plonsky & Oswald, 2014, for effect size interpretation in the neighboring field of L2 research). One clear downside of this approach is that these cutoffs are not topicspecific (or even field-specific). ...

Reference:

What if we want to test for similarities between groups, rather than differences? Equivalence testing techniques for corpus linguistics