Abstract—Database broadcast is an effective and scalable approach to disseminate information of high affinity to a large collection of mobile clients. A common problem of existing broadcast approaches is the lack of knowledge for a client to determine if all data items satisfying its query could be obtained from the broadcast. We therefore propose a semantic-based broadcast approach. A semantic descriptor is attached to each broadcast unit, called a data chunk. This semantic descriptor allows a client to determine if a query can be answered entirely based on broadcast items and, if needed, identify the precise definition of the remaining items in the form of a “supplementary” query. Data chunks can be of static or dynamic sizes and organized hierarchically. Their boundary can be determined on-the-fly, adaptive to the nature of client queries. We investigate different ways of organizing the data chunks over a broadcast channel to improve access performance. We introduce the data affinity index metric, which more accurately reflects client-perceived performance. A simulation model is built to evaluate our semantic-based broadcast schemes.