Recent years have witnessed an increasing amount of research efforts and applications on keyword bid in search engine advertising, giving rise to a focal point of research for the question of how to generate more relevant and novel keywords in targeting users’ intents. This paper introduces the notion of keywords with indirect relevance based on query logs, and then proposes a new semantic transfer keyword suggestion method (namely, STK) to generate the candidate keywords with links to the seed keyword in light of semantic transfer paths that are formulated with multi-step relevance. To reduce computational consumption in generating the links, a pruning strategy is developed based upon certain properties of semantic transfer paths, so that those unnecessary links are cut in the algorithmic process. Furthermore, the STK method is tested via data experiments and illustrative examples, demonstrating the overall superiority of STK in comparison with other major keyword suggestion methods of concern for effectiveness measures.