Web service search has been a serious concern for service-oriented software development. The challenge is how to find the right Web services efficiently and effectively for a given development task. There have been many efforts on developing techniques or systems to search services. Though effective in certain scenarios, existing techniques do not form systems for public use; or are based on one query modal--keyword query, which cannot give the matched services accurately. In this paper, we introduce proposed multimodal query search, where users can use keyword and file as query or custom the query. The multimodal query is based on an innovative similarity measure approach, which incorporates both semantic information and structural information of Web services. Our experiments and test cases validate the effectiveness of the approach. Compared with the alternative system Seekda, it is able to obtain much higher search accuracy with keyword query (with a match rate of 2-4 times higher than that of Seekda). The custom search can achieve 100% top-3 match rate, while Seekda fails in most cases using keywords.