On-demand Service Discovery Driven By Domain-specific Goal Knowledge

With the rapid development of service-oriented architecture(SOA)and its related techniques,a lot of Web services have been created and used for developing software systems.This service-oriented software development(SOSD)paradigm can not only reduce the time and cost of development,but also increase the quality and flexibility of software systems.Although service-based software systems are becoming popular,SOSD still faces many issues.Particularly in the requirement engineering phase,it is difficult to help users achieve on-demand service discovery due to the lack of effective knowledge management strategy.A large amount of related works on service knowledge management and service discovery have been carried out at home and aborad.As for service knowledge management,existing researches mainly involve three aspects:service-oriented ontology learning,service composition pattern mining,and goal-oriented service requirement modeling,while little work has been devoted to obtaining goal knowledge that can represent services' functionalities from service resources(especially natural language textual descriptions).As for service discovery,most existing researches leverage interfaces in structural descriptions(e.g.,WSDL),while limited attentions have been given to the business functional goals included in textual descriptions.In addition,although many effective logic-based semantics-aware service discovery approaches have been proposed,they are hard to apply.Current service registries still mainly adopt keyword matching-based service discovery approaches,which suffer from low performace and is insufficient to satisfy user requirements.Moreover,existing service discovery researches have no effective method for handling the vague expression of user requirements.Vague user queries will affect the results of service discovery.Towards the issues about existing researches on service knowledge management and service discovery,this thesis attempts to address the following key problem:"how to achieve on-demand service discovery driven by domain-specific goal knowledge".Considering the key problem,the research works conducted in this thesis can be summarized as follows.(1)A functional goal extraction approach is proposed for textual descriptions of services.It uses the Stanford Parser to parse sentences contained in services' textual description to obtain two types of grammatical information:POS tagging and Stanford dependencies.Through in-depth analysis of grammatical information of sentences with different structures,an algorithm is designed for functional goal extraction from grammatical information of a sentence.Considering services' domain characteristics,the algorithm refines extracted candidate functional goals by using importance information of the domain to which a service belongs,so as to obtain functional goals with high domain representation.(2)A domain-oriented functional goal aggregation approach is proposed to identify commonality and variability of similar(including semantically similar)functional goals in a specific domain.It first makes statistics of functional goals extracted for all services in a domain.Semantically similar functional goals are then transformed using synonymous verb groups and synonymous noun groups that are built based on the WordNet.Finally,three algorithms are designed for aggregating functional goals.(3)A goal-based service discovery approach is proposed by leveraging functional goals extracted from services1 textual descriptions.In reality,most users are usually not equipped with enough domain background knowledge and lack understanding of service functionalities in existing service registries,it is difficult for them to specify queries that can accurately reflect their functional requirements,which will in turn affect the results of service discovery.To solve this issue,after clustering the functional goals of each domain in a service registry,a goal-oriented query expansion approach is designed.Given an initial query,semantically similar functional goals of the query will be recommended,and the user can gain prompt understanding of service functionalities relevant to his/her functional requirements and then select satisfactory functional goals as a new query.Afterwards,a goal matching-based service retrieval approach is designed by comparing functional goals of each candidate service with those selected by the user,which can obtain accurate services that can satisfy user requirements.Moreover,considering the limitation of the proposed functional goal extraction approach,a hybrid service discovery approach is designed by integrating the goal matching-based service discovery approach with existing service discovery approaches,in order to improve the recall performance.We validate the effectiveness of approaches proposed in this thesis by conducting a series of experiments on a real-world Web API service dataset crawled from a popular service registry,ProgrammableWeb.