搜索广告系统简单架构

Query Analysis:

纠错、同义变换、改写

切词、提取主题词

意图分析

User Server:

Demographic 

Cookie

Session

Board match server:

实现宽泛匹配

宽泛匹配的相关性很重要

e.g.  Query Rewriting:

rewrite user query q into Q' = (q1,q2,q3,...)

Use EM to select ads for Q'

Fits well in the current system architectures

Index Server:

实现精确匹配、短语匹配

各种业务过滤功能

查找广告倒排索引

Ad Server:

正排表查找广告物料：创意标题、描述、URL

CTR Server:

实现CTR预估功能

计算密集，内存密集

Auction Server:

广告排序:Q*Bid

广告计价 Price = Q'*Bid'/Q