The rapidly growing amount of available digital documents of various formats and the possibility to access these through internet-based technologies in distributed environments, have led to the necessity to develop solid methods to properly organize and structure documents in large digital libraries and repositories. Specifically, the extremely large size of document collections make it impossible to manually organize such documents. Additionally, most of the documents exist in an unstructured form and do not follow any schemas. Therefore, research efforts in this direction are being dedicated to automatically infer structure and schemas. This is essential in order to better organize huge collections as well as to effectively and efficiently retrieve documents in heterogeneous domains in networked system. This paper presents a survey of the state-of-the-art methods for inferring structure from documents and schemas in networked environments. The survey is organized around the most important application domains, namely, bio-informatics, sensor networks, social networks, P2P systems, automation and control, transportation and privacypreserving for which we analyze the recent developments on dealing with unstructured data in such domains.