A semantic transfer approach to keyword suggestion for search engine advertising

Search Engine Advertising has been widely adopted by advertisers to target potential consumers. However, the advertisers generally focus on limited popular advertising keywords, leading to fierce competition. Therefore, abundant relevant keywords need to be discovered to reduce the advertising cost. In this regard, this paper proposes a novel semantic transfer approach (named STAKS) to suggesting keyword for search engine advertising. Compared with the existing methods which explore keywords with direct relevance to the given seed keyword, STAKS can find keywords with multi-step indirect relevance through semantic paths. Moreover, three pruning strategies are designed to (1) ensure the relevance between the suggested keywords and the seed keywords, (2) narrow the semantic drift and (3) reduce the computational consumption. Data experiments show the superiority of STAKS which finds more novel keywords, owing to the indirect relevance ignored by existing methods. Therefore, STAKS is deemed effective in supporting the advertisers to achieve high advertising impressions with relatively low bidding prices.