4. agenda Content match basics Ad selection Ad selection basics Finding advertising keywords on web pages Holistic view at the page in Contextual Advertising Using click data to improve IR ad retrieval When to advertise (relevance threshold) the complete pipeline
5.
6.
7. 谷歌 2003年3月, google发布adsense 2004年10月,进入中国 AdSense for Feeds AdSense for search AdSense for mobile content AdSense for domains AdSense for video 雅虎 2003年6月,Overture公司发布CM广告(Content Match); 雅虎YPN(Yahoo Publisher Network) 2007年, 推出新一代CM平台keystone 微软 2006年,微软推出广告计划adcenter 百度 2005年,百度推出主题推广CPRO 窄告, 2004年推出 阿里妈妈,2007年推出
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18. agenda Content match basics Ad selection Ad selection basics Finding advertising keywords on web pages Holistic view at the page in Contextual Advertising Using click data to improve IR ad retrieval When to advertise (relevance threshold) the complete pipeline
19. Ad selection basics Objective Database approach and IR methods Understanding the pages, understanding the ads, matching A few considerations
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40. agenda Content match basics Ad selection Ad selection basics Finding advertising keywords on web pages Holistic view at the page in Contextual Advertising Using click data to improve IR ad retrieval When to advertise (relevance threshold) the complete pipeline
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52. agenda Content match basics Ad selection Ad selection basics Finding advertising keywords on web pages Holistic view at the page in Contextual Advertising Using click data to improve IR ad retrieval When to advertise (relevance threshold) the complete pipeline
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
68. agenda Content match basics Ad selection Ad selection basics Finding advertising keywords on web pages Holistic view at the page in Contextual Advertising Using click data to improve IR ad retrieval When to advertise (relevance threshold) the complete pipeline
70. Why CTR model/prediction Traditional IR is not enough Foundation for ad serving optimization Building block for CPM calculation Quality score
71.
72. CTR Modeling/Prediction Will the user click on the ad at exactly this time at this page? CTR(u, a, p) ,click-through rate, very small Separate into three parts: Feature selection (first extract, then select) Model building Evaluation metrics Focusing on features Scalable model building is key
74. agenda Content match basics Ad selection Ad selection basics Finding advertising keywords on web pages Holistic view at the page in Contextual Advertising Using click data to improve IR ad retrieval When to advertise (relevance threshold) the complete pipeline
75.
76.
77.
78.
79.
80. agenda Content match basics Ad selection Ad selection basics Finding advertising keywords on web pages Holistic view at the page in Contextual Advertising Using click data to improve IR ad retrieval When to advertise (relevance threshold) the complete pipeline
81. The complete pipeline Ad selection and pricing Serving pipe and building pipe Request understanding, ads understanding, matching Trigger, rank1, rank2, rank3… ……