We are talking about verbal Behavior, not about something inherent in an arrangement of letters or sounds. Skinner pointed out that the 'meaning of a word' really refers to the situational context - when, where, under what circumstances - it is or isn't Used. These behaviors are learned and trained over a period of time by a particular verbal community, in more or less mutual agreement. Like any other behavior there are situations where it is highly predictable that it (the behavior or word) will (or won't) occur, and others where it is more iffy and probabilistic. Some iffy examples for 'chair' might be stool, sofa, bench, trapese, swing, sling, hospital bed; or among other things one can sit on, ledge, lap, hands, face, fence, side lines.
Generalization & discrimination overlap as in other stimulus contexts (is this wavelength (which isn't a 'word') REALLY 'blue' or 'green'?, this pH 'sour'?). Asking if a stool is a chair is like asking if gray is Really white, or black. Or put otherwise, if black (chair) includes gray (stool).
Heisenberg would smile.
Ambiguity is acceptable, being unavoidable.