Summarization based on GenSpace graphs aggregates data into summaries in many ways and identifies summaries that are far from user expectations. Mining interesting summaries in GenSpace graphs involves expectation propagation and interestingness measure calculation in the graphs. Both the propagation and the calculation need to traverse the GenSpace graph, but the number of the nodes in the GenSpace graph is exponential in the number of attributes. In this paper, we propose pruning methods in the different steps of the mining process: pruning nodes in ExGen graphs before constructing the GenSpace, pruning nodes in GenSpaces before propagation, and pruning nodes in GenSpaces during propagation. With these methods we make the traverse more efficient, by reducing the number of the nodes visited and the number of records scanned in the nodes. We also present experimental results on the Saskatchewan weather data set.