In recent times, the proliferation of spatial data on the Internet is beginning to allow a much larger audience to access and share data currently available in various Geographic Information Systems (GISs). Unfortunately, even if the user can potentially access a huge amount of data, often, she has not enough knowledge about the spatial domain she wants to query, resulting in a reduction of the quality of the query results. This aspect is even more relevant in integration architectures, where the user often specifies a global query over a global schema, without having knowledge about the specific local schemas over which the query has to be executed. In order to overcome such problem, a possible solution is to introduce some mechanism of query relaxation, by which approximated answers are returned to the user. In this paper, we consider the relaxation problem for spatial topological queries. In particular, we present some relaxed topological predicates and we show in which application contexts they can be significantly used. In order to make such predicates effectively usable, we discuss how GQuery, an XML-based spatial query language, can be extended to support similarity-based queries through the proposed operators.