Information awareness and knowledge discovery has become one of the key issues currently in distributed, heterogeneous and massive software development. In this situation, semantic Web was introduced into software engineering to build fine-grain semantic links between multi-source heterogeneous data. And a novel approach was proposed to build ontology, extract and recover links, and further construct ontology-based software engineering linked data automatically. It extracted and merged ontology concepts, resolved entities and their properties, and built complete linked data without redundancy from structural data sets in software repository. Also it recovered missing linked data from software repository using Natural Language Processing (NLP) and Information Retrieval (IR) techniques with three features including synonym, verb-object phrase and structural information. The experimental results show that the proposed approach can construct and merge software engineering ontology automatically from distributed software engineering data sets, recover missing linked data and enlarge ontology effectively. Compared with Baseline, Phrasing and O-CSTI, this approach performs much better in recall, precision and F-measure.