From generic to specific: A semantic network of search keywords

When paid search advertising has become a predominant form of online advertising, how to select appropriate keywords becomes a meaningful issue for advertisers. Bidding for keywords which are semantically similar with hot keywords is one effective way for advertisers. However, due to the difficulty to guess similar keywords, advertisers in reality tend to bid for a few expensive but hot keywords. While no study in marketing can solve the problem of finding similar keywords empirically, this paper builds a semantic network of search keywords where similar keywords are subsumed in the same keyword groups. Further, we find that keyword genericalness has a positive effect on keyword centralities in the network, and keyword centralities can improve their impressions and clicks, but result in a lower click-through rate.