Hierarchical Bayesian-based Keyword Generation And Selection In Sponsored Search Advertising

In the past ten years,sponsored search advertising has developed into the mainstream business model of general search engines(such as Google or Baidu),and it also occupies an important position in e-commerce platforms.The current research on sponsored search advertising focuses on bidding prices and budgeting strategies.There are relatively few studies on the generation,expansion and selection of keywords,and the existing research is insufficient.First,most of the existing keyword generation and expansion methods obtain popular keywords,which are not cost-effective.Secondly,the existing method generates a small number of keywords,which is difficult to meet the needs of advertisers with sufficient budget.Finally,the existing keyword selection methods are only for a single advertising platform,and advertisers often choose multiple platforms to advertise at the same time in real work.The research of this article proposes the optimization strategy of keyword generation,expansion and selection in sponsored search advertising.First,this dissertation proposes a keyword generation method based on hierarchical Bayes.With only a few seed keywords provided by advertisers,the keyword generation method proposed in this article adopts Wikipedia web pages as the source text corpus,and uses hierarchical Bayesian method to estimate the model parameters,then generates keywords by taking advantage of the rich link structure and page content of Wikipedia.The method is evaluated in the experiment from three aspects of relevance,professionalism and cost performance.The result proves the advantages of this method.Second,this dissertation uses the link relationship of the network hierarchy between Wikipedia pages to propose a new keyword expansion strategy called WIKG.WIKG explores the link structure of Wikipedia categories,and uses it as a basis to construct a network graph of entry articles in an iterative manner to achieve flexible expansion of keywords.The termination condition is determined by a threshold reflecting the tradeoff between coverage of the generated keyword set and its relevance to seed keywords.Experimental results show that the WIKG outperforms three baselines derived from the extant literature,in terms of both coverage and relevance.Third,this dissertation proposes a multi-platform keyword selection strategy.User exposure to repeated advertising from multiple sources can increase sales,which requires consideration of multi-platform factors in keyword selection.The non-parametric hierarchical Bayesian keyword CTR(NHBKC)model proposed in this dissertation is aimed at advertisers facing the actual situation of multiple advertising platforms,extracting cross-platform multi-dimensional user portrait feature tags.This article estimates the comprehensive CTR value of each keyword based on the advertising click behavior data of the segmented users by NHBKC.Then a keyword selection strategy is proposed based on the estimated results of the NHBKC model,and the reconciliation parameters are introduced to divide the performance of the keyword selection into multiple parts for calculation.The parameters are adjusted to achieve a balance between brand promotion and profit to meet the different requirements of different advertisers and the same advertiser in different time periods for advertising goals.The keyword selection method in this article reserves space for advertisers to adjust.Advertisers can easily add,delete or change user feature tags according to changes in consumer group characteristics.The internal division of each feature tag can also be adjusted at any time.It is convenient for advertisers to adjust keyword selection and advertising promotion plan at any time to keep up with the pace of market changes.The real data of service providers were used in the experiment.Experimental results show that this method can help advertisers achieve better results in keyword selection.In short,the keywords generated by the hierarchical Bayesian keyword generation and expansion method proposed in this dissertation can strike a balance between the coverage and relevance of the advertiser's products(or services)and the niche market in which they operate.The methods in this article provide more keywords,and most of them are long-tail keywords that meet the requirements of advertisers.The methods have advantages in coverage,relevance,and professionalism.The keyword selection strategy in this article considers the actual situation of advertisers facing multiple advertising platforms.It can balance the two major advertising goals of brand promotion and profit according to different advertisers and the different requirements of the same advertiser at different times.The strategy reserves space for advertisers to adjust,so that advertisers can adjust their keyword selection and advertising promotion plans in accordance with market changes at any time.