In order to improve software dependability, a method is proposed for solving technical contradictions in dependable software development based on quality function deployment (QFD) and the Russian theory of inventive problem solving (TRIZ). First, a dependable software planning house of quality (HOQ) is built to introduce the voice of customers into the software development process. Then, the technical characteristic pairs whose relationships are negative in the technical correlation matrix are found out and analyzed in detail. The inventive principles of TRIZ are applied to solve the technical contradictions, and several feasible solutions or inspirations could be acquired. Considering the vagueness and uncertainty of expert judgment, and the case of multi-granularity and multi-semantics in an expert’s linguistic statement, the linguistic information decision-making theory developed in recent years is introduced to evaluate and select the best feasible alternative. Finally, a real-world case of astronautics of an embedded real time operating system (ERTOS) software development is provided to demonstrate the validity and applicability of the proposed method.