A major challenge with requirements engineering is that the requirements to be captured usually are described in qualitative terms which are imprecise in nature. However, most of the existing software specification methodologies require that the requirements be represented in a precise form. In this paper, we propose 1) to use soft functional requirements in our task-based specification methodology (TBSM) to capture the imprecise requirements, and 2) to formulate soft functional requirements using fuzzy logic. More specifically, the soft conditions are represented using the canonical form in test-score semantics. The proposed approach not only can facilitate the exploration of various trade-offs during the design phase, but also can enable a more realistic validation of the user's imprecise requirements.