Presentation is loading. Please wait.

Presentation is loading. Please wait.

Output URL Bidding Panagiotis Papadimitriou, Hector Garcia-Molina, (Stanford University) Ali Dasdan, Santanu Kolay (Ebay Inc) Related papers: VLDB 2011,

Similar presentations


Presentation on theme: "Output URL Bidding Panagiotis Papadimitriou, Hector Garcia-Molina, (Stanford University) Ali Dasdan, Santanu Kolay (Ebay Inc) Related papers: VLDB 2011,"— Presentation transcript:

1 Output URL Bidding Panagiotis Papadimitriou, Hector Garcia-Molina, (Stanford University) Ali Dasdan, Santanu Kolay (Ebay Inc) Related papers: VLDB 2011, InfoLab TR-939, AdAuctions 2009

2 Search Engine Results Page (SERP) Organic Results Sponsored Ads Query Sponsored Search Ads

3 Keyword Bidding Advertiser Search Engines the social network lord of the rings the matrix lotr III... # keywords = ~ 10K KEYWORDS

4 Example SERPs en.wikipedia.org/wiki/The_Social_Network www.imdb.com/title/tt1285016/ www.imdb.com/title/tt133093/ en.wikipedia.org/wiki/The_Matrix en.wikipedia.org/wiki/The_Lord_of_the_rings www.imdb.com/title/tt167260/ www.imdb.com/title/tt120737/ the social network the matrix the lord of the rings lotr iii

5 Output Bidding Advertiser Search Engines imdb.com AND wikipedia.org # URLs = 2 URLs

6 Outline Architectures Bid Language Output bid/expression generation Spill Evaluation Experiments

7 Architectures Current Search Engine Architecture

8 Architectures Serialization Overview – First, retrieve organic results – Then, retrieve ads Pros – Simplicity Cons – Results Latency O: Organic Search System S: Sponsored Search System SERP

9 Architectures Pipelining Split organic search system to – Or: retrieval subsystem (retrieve relevant docs) – Op: post-processing subsystem (create result snippets) Op and S run in parallel Pros – No additional latency Cons – Sponsored search system depends on organic system O: Organic Search System = Or + Op S: Sponsored Search System SERP

10 Architectures Parallelization URLs with ads are known a priori S can use – Or’: Or replica that indexes only URLs with ads Pros – No additional latency – Independent organic and sponsored search system Cons – More resources O: Organic Search System (Or + Op) S: Sponsored Search System Or’: Small replica of Or V: Ad validation SERP

11 Bid Language Model Output Expression – e.g., a := (u 1  u 2 )  u 3  (h 1  h 2 ) – u: URL e.g., en.wikipedia.org/wiki/The_Social_Network – h: host e.g., en.wikipedia.org Questions – URLs or hosts or both? – complex or simple?

12 Output Expression Generation Motivation Use existing keyword campaigns to generate realistic output expressions to study The social network lord of the rings the matrix lotr III … Output Expression Generator imdb.com AND wikipedia.org

13 Problem – INPUT: keyword set R – OUTPUT: expression a that “covers” R Candidate solutions – a 1 := u 1  u 2  u 3 – a 2 := u 1  u 4 – a 3 := u 5 Output Expression Generation Motivating Example

14 Compactness Contain few URLs Spill minimization: Do not match “irrelevant” queries Output Expression Generation Objectives Output Expression Size |a| Spill spill(a,R) a 1 := u 1  u 2  u 3 3{} a 2 := u 1  u 4 2{q 5 } a 3 := u 5 1{q 4,q 5, q 6 }

15 Query Set Output Cover minimize γ|a| + (1-γ) |spill(a, R)| subj. to m(a,q),  q  R γ : regularization parameter Related to – Set Cover – Red-Blue Set Cover Output Expression Generation Problem Statement

16 Output Expression Generation Greedy Algorithm Pre-compute – C[u]: Queries covered by URL u – S[u]: Spill of URL u w.r.t. R Algorithm

17 Spill Evaluation Spill queries may be relevant to R Divide spill(a, R) to – positive: relevant – negative: irrelevant Use query clustering for evaluation Example: – a := u 2  u 3 – Positive spill = {q 1 } – Negative spill = {q 5 }

18 Experimental Evaluation Goals Compare output URL bidding variations – 1-URL, 2-URL, 3-URL e.g, 2-URL: use only URLs, up to 2 URLs in a disjunct – 1-host, 2-host, 3-host – 1-mixed, 2-mixed Comparison criteria – Compactness – Spill tradeoff – Spill Evaluation

19 Experimental Evaluation Setup Dataset (from Yahoo query logs) – 12,931,117 queries – 62,666,514 URLs – 7,185,392 hosts – 2,251 ads Process – For each variation (1-URL, 2-URL, …) For different γ values – Generate output expressions for all 2,251 ads

20 Experimental Evaluation Compactness vs Spill

21 Experimental Evaluation Positive and Negative Spill

22 Experimental Evaluation Summary Compactness-spill trade-off – Using both URLs/hosts outperform other options – Up to 2 conjuncts in a disjunct is sufficient Spill evaluation – Output expressions can bring additional queries Other experiments in Combining keyword and output bidding – Output expression are suitable for half of the keywords – Using only hosts seems to be sufficient

23 Conclusions Output URL bidding can be implemented efficiently Advantages over keyword bidding – Bid Compactness – More relevant queries

24 THANK YOU!


Download ppt "Output URL Bidding Panagiotis Papadimitriou, Hector Garcia-Molina, (Stanford University) Ali Dasdan, Santanu Kolay (Ebay Inc) Related papers: VLDB 2011,"

Similar presentations


Ads by Google