With the advanced database technology developed during the past decades, it is possible to store a vast amount of information in computers. This explosive growth of information in databases has generated an urgent need for new techniques and tools that can intelligently and automatically transform the stored data into useful information and knowledge. In this paper, a novel Hybrid Knowledge-Based Connectionist Network (HKBCN) for knowledge discovery is presented. In the HKBCN framework, initial domain knowledge is firstly embedded into a connectionist network structure. Then, this primitive structure evolves to minimize empirical errors. HKBCN has the ability of transferring the role of learning into knowledge refinement and generate rule-based explanations. An example is provided in the area of natural language processing (NLP) to illustrate the way in which HKBCN operates.