中文摘要：

针对海量的信息提供类（InformationProviding）语义Web服务的自动组合问题，提出了一种将正向搜索与逆向搜索相结合并追求QoS最优的快速组合方法．该方法首先根据用户的请求对服务库中的服务进行层次划分，去除无关服务并得到具有并行调用结构的组合服务原型即规划图；之后，利用得到的参数的服务源信息，计算每个服务的最优组合QoS最后，以用户请求的输出参数为起点进行逆向搜索，以最优组合QoS为约束条件构造所有可能的最优解．以WS—Challenge2009提供的竞赛集进行实验，结果表明该方法不仅能够找到组合服务的最优QoS，而且可以获得大量结构不同但OoS均为最优的解，从而为组合服务的执行提供更多的选择．

英文摘要：

Targeting the automatic composition problem of large scale of semantic information providing Web services, this paper proposes a novel approach to find the solutions with the opti- mal global QoS. First, it separates services into ordered layers according to the user request and removes the redundant services to get a parallel composition structure （i. e. , a planning graph）. Then, it computes the optimal local QoS values according to the parameter sources which are services producing that parameter. Finally, it searches all possible optimal solutions from the us- er requested outputs. Experiments on the test sets offered by the WS-Challenge 2009 show that our approach cannot only find the optimal global QoS value but also can get all the optimal solu- tions with different composition structures to provide more options for the service execution.