Automatic Construction Of Crowded Software Engineering Linked Data Based On Ontology

Crowded software engineering is becoming a new model for software development in the cloud era.With the help of community development and swarm intelligence technique,crowded software engineering could quickly and effectively construct software which has large scale,complex functions and technology innovation.Unfortunately,cooperative development is facing the challenge of mass data.In crowded software projects,there are massive developers,codes,versions,requirements,design models,test cases,bugs,commits,tasks,discussion records,mail lists,etc.How to effectively aware information and discover knowledge from the large-scale distributed data has become an urgent problem.To resolve this problem,this paper applies the semantic web technique to software engineering,and links the multi-source heterogeneous data at fine-grained semantic level.Based on ontology technique,this paper automatically constructs software engineering linked data among requirements,modules,codes,bugs,commits,developers,mail links,etc.Using these linked data,crowded software engineering activities including information intelligent search,data mining and change impact analysis becomes more effective.This paper makes the following major contributions:1)Software engineering ontology construction.With mapping-based technique,this paper constructs original ontologies from structured metadata in different data sources,then merges these original ontologies to construct software engineering ontology.In this paper,mapping-based original ontology construction technique is improved by extracting and merging relational schema,and ontology merging technique is improved by using name,property,relation and structure to identify similar ontology concepts.Comparing with related techniques using three open source projects including eclipse.jdt,tomcat7 and openssh,the experimental results show the effectiveness of this proposed method.2)Software engineering linked data extraction.This paper uses mapping-based technique to extract software engineering linked data from structured data in different data sources,then cleans data with entity resolution technique and property elimination technique to decrease redundancy and conflict.The experiment based on open source projects shows this proposed method will identify and process redundancy and conflict in linked data effectively.3)Software engineering linked data recovery.Based on three features including synonyms,verb-object phrases and structural information,this paper uses natural language processing technique and information retrieval technique to recover potential and missing linked data from structured and unstructured data.Using the data sets from two open source projects and two closed source projects to make experiments for method comparison and features comparison,the results show that both the precision and the recall of this proposed method are better than others in recovering linked data,and verb-object phrases is the most effective feature to recover linked data.