With the growing amount of XML and RDF data, keyword search over XML and RDF data has become an important and increasingly researched topic. In this thesis, we investigate several problems related to keyword search over XML and RDF data, and provide solutions to these problems. The research consists largely of three main technical contributions: top-k nearest keyword (NK) search for XML data, query suggestion for XML data and query suggestion for RDF data. We first study the top-k NK search problem for XML data, which provides an approach to exploring XML queries by the distance between keyword matching nodes. A top-k NK query is to find the top-k nearest neighbors of a given node where each neighbor matches a certain keyword, and it can serve as the building block to deal with many problems in XML data, such as keyword search. In our research, we design a method to construct an extended compact TVP (ecTVP) index to efficiently find the top-k nearest neighbors in XML data. We build the index by constructing a variant of Extended Compact Tree, which finds the top-k nearest neighbors during a bottom-up and a top-down process. Theoretical analysis and experiments indicate that our proposed method is efficient to build the ecTVP index. Moreover, we reduce the redundancy in the ecTVP index. In this way, the index costs less space and query time.