Abstract
This chapter explores a new methodology for extracting multi-word units that were once common but have since become obsolete from large corpora (esp. from the Google ngrams dataset of the Google Books project). It complements a modified frequency-based methodology previously used for detecting lexical obsolescence (Tichý, 2018) with a bottom up approach to calculating association measures in multi-word sequences inspired by Wahl & Gries (2019).
The analytical part examines expressions identified as potentially obsolete on their way from Late Modern to Present-day English. Conditions, circumstances and consequences of the loss of such expressions are considered with a focus on the competing forms expressing similar functions that may be recognized as supplanting the old forms.