Web of Things applications require advanced solutions to provide adaptation to different purposes from common context models. While such models are application-specific, the adaptation itself is based on questions (i.e. concerns) that are orthogonal to application domains. We rely on semantic technologies and reason about contextual information to evaluate, in near real-time, the pertinence of each adaptation possibility to adaptation questions covering various concerns. We evaluate our solution against a smart agriculture scenario using the ASAWoO platform, and discuss how to design context models and rules from ``classical'' information sources (e.g. domain experts, device QoS, user preferences).