An Approach for Semantic Web Discovery Using Unsupervised Learning Algorithms

Abstract

With the huge amount of available web services, it becomes increasingly difficult to find target web services for users accurately and effectively. For this reason, research in web service clustering has recently gained much attention. Most existing clustering methods perform well when dealing with long text documents. However, the textdescriptions of web services are in the form of short text. Meanwhile, it is meaningful to consider word order information in the textdescriptions of web services. Hence, we presented a service discovery approach based on web service clustering considering this issue. In our approach, web service discovery was divided into two parts: web service clustering and web service selection. In the process of web service clustering, the textdescriptions of web services were represented as vectors. In order to make vector representations reflect as much as possible the semantic information contained in the web service textdescriptions, we tried four different unsupervised sentence representations. In another part, LDA was used to mine topic semantic information of web services after user’s web request was placed into a specific cluster according to its web service textdescription vector. The final efficiency of web service discovery was used to measure the effectiveness of our approach.