This paper proposes a self-adaptive mobile web service (MWS) discovery framework for a dynamic mobile environment (DME) to deal with MWS proliferation, dynamic context, and irrelevant MWS discovery challenges. The main contribution of this research includes an improvement of the matchmaking algorithm, enhanced MWS categorization approach, and extensible meta-context ontology that represents the context information in DME. This was achieved by enabling the self-adaptive matchmaker to learn MWS relevance using a Modified-Negative Selection Algorithm (M-NSA) and retrieve the most relevant MWS based on the current context of the discovery. To assess the proposed framework, series of experiments was carried out using publicly-available datasets. The performance of the framework is evaluated against the state-of-the-art frameworks. It was found that the proposed framework is more effective and attained better binary and graded relevance when subjected to context variations which are prevalent in DME. This is useful for service-based application designers and other MWS clients.