“Works on my machine”: A case study of replicability challenges in computational humanities research