An efficient term mining method to build a general term network is presented for entity relation visualization and exploration. Terms from each document in the corpus are first identified. They are subjected to an analysis for their association weights, which are accumulated over all the documents for each term pair. The resulting term association matrix is used to build a general term network. A set of terms having similar attributes can then be given to extract the desired sub-network from the general term network for visualization. This analysis scenario based on the collective terms of the similar type or from the same source enables evidence-based relation exploration. Some practical instances of crime investigations were demonstrated. Our application examples show that term relations, be it causality, coupling, or others, can be effectively revealed by our method and verified by the underlying corpus. This work contributes by presenting an efficient and effective term-relationship mining method and extending the applicability of term networks to a broader range of informatic tasks.