Research On Relevance Technology Based On Multi-Perspective Fusion Of Search Advertisement

Search advertising refers to a specific application of the advertising system.Displaying the related advertising products to the user according to the user searching words.With the continuous development of search engine,search advertising has become more and more important in the advertising system and its economic benefit is much higher than other advertisements.Search advertising relevance refers to the relevance of the user searching word with advertising products.Since search advertising is based on the user searching word,it is extremely important to study the meaning of this relevance.Whether it is possible to recommend high relevant product advertisements to users and make users more satisfied with the search result page has become an important target of search engines.This thesis summarizes the current research status of search engine and recommendation system,as well as relevance research methods at home and abroad.This thesis puts forward a multi-perspective fusion of search advertising relevance method so as to solve the specific relevance of the problem combined with the business needs of the PDD.The system of the high relevance designed and implemented in this thesis is seen as part of the application layer of the search advertising platform.The relevance system ensures the relevance of the company and improves the user experience.For the relevance system,this thesis solves the relevance problem from the design of data processing scheme,the design of multi-perspective fusion algorithm and the design of relevance service.The relevance research of multi-perspective fusion proposed in this thesis includes the perspective of text relevance,the perspective of extended recommendation relevance of advertising products,the perspective of ranking relevance of product attributes and the perspective of user intention recognition relevance.This thesis designs and realizes each angle one by one,so as to realize the relevance of the final multi-perspective fusion.The relevance study of multi-view fusion in this thesis combines the application of neural network and depth model on the basis of traditional machine learning model.At present,the project has been officially launched.The plan of this thesis for the PDD search advertising platform increases the click-through rate about 1.23%,increases the consumption about 1.44% and increases the spam about 40%.