In many application domains (e.g., WWW mining, molecular biology), large string datasets are available and yet under-exploited. The inductive database framework assumes that both such datasets and the various patterns holding within them might be queryable. In this setting, queries which return patterns are called inductive queries and solving them is one of the core research topics for data mining. Indeed, constraint-based mining techniques on string datasets have been studied extensively. Efficient algorithms enable to compute complete collections of patterns (e.g., substrings) which satisfy conjunctions of monotonic and/or anti-monotonic constraints in large datasets (e.g., conjunctions of minimal and maximal support constraints). We consider that fault-tolerance and softness are extremely important issues for tackling real-life data analysis. We address some of the open problems when evaluating soft-support constraint which implies the computations of pattern soft-occurrences instead of the classical exact matching ones. Solving efficiently soft-support constraints is challenging since it prevents from the clever use of monotonicity properties. We describe our proposal and we provide an experimental validation on real-life clickstream data which confirms the added value of this approach.