Data-services are applications whose main concern is to provide data to theirclient applications. Data-services play a key role in areas like the Internet ofThings (IoT), where smart objects might want to offer/consume data throughInternet, and thus, providing/discovering such data-services automatically.To make data-services discoverable, the usual strategy is to register dataservicesin some kind of service-broker, i.e., a marketplace where data-services arepublicly offered. Then, smart objects query the service-broker, and the servicebroker is responsible to match the request with its data-services. How to perform this matching automatically is still an open problem in IoT.In this paper, we propose a framework for specifying data-services so thatthey can be automatically discovered. To achieve it, we provide unambiguousdescriptions of the data-services and the request, together with a mechanismcapable of interpreting these descriptions and check whether they match. Oursolution is grounded on ontology-based data integration and can be applied inthe IoT context, altought it can also be used in any other domain involving thediscovery of applications retrieving data.In essence, our idea is based on, given a domain ontology describing thereal world our data-services speaks about, consider each data-service as a newassociation in that ontology. Indeed, a data-service consuming some input objectsand retrieving some output objects can be modelled as an association from theproper to the latter. As expected, ontology constraints must be used to restrictthe instances of the association to the input-output instances our data-serviceexpects/provides.Hence, the problem of matching data-services is reduced to that of automaticreasoning on ontologies (and in particular, association subsumption). Thus, contributions on this last field can be directly used for the data-services discovering problem.