Taxonomy learning for semantic annotation of web services

Web Service composition offers the facility to create new services satisfying a certain functionality which can not be assured by a single existent service. The full potential of Web Service composition can only be achieved if the process is automated. Services should first be annotated with semantic information in order to do automatic service composition. The semantic information is provided by ontologies. This paper presents an unsupervised approach to automatically build a domain specific taxonomy from textual descriptions. The approach is based on hierarchical self-organizing maps. The candidates for concept names are collected by mining text corpora. The term extraction process is based on recognizing linguistic patterns in a text corpus. The taxonomy has been built in the framework of the Food Trace project [20] for traceability in the domain of food industry. The leamt taxonomy represents the domain specific branches of an ontology used for semantic annotation of the web services. Taxonomy concepts are semantic descriptions of the inputs and outputs of the operations provided by a Web Service.