The reference that how checks search engine measures volume (before piece)
Author: Engineer of United States of Yahoo of old morning sunlight
Cultural background: The quality index that searchs engine includes dependency commonly (Relevance) , sex of effectiveness for a given period of time (Freshness) , comprehensive sex (Comprehensiveness) with usability (Usability) wait for 4 respects, the index that we should talk about today measures the category that belongs to integrality index.
What need to pay attention to above all is, to searching engine, the index volume of the webpage and capture quantity are different idea. The webpage capture amount that searchs engine wants ambitious Yu Suo to bring an amount commonly, because very much content includes in the webpage of capture,repeat or cogged wait for the webpage with not high quality. Algorithm takes ground of search engine need among the webpage from capture its elite, go its scum, choose a valuable webpage to have index. Accordingly, to the user character, the index that searchs engine measures volume ability is more significant.
Next, increase index to measure without limit and can not assure to search the promotion of quality surely. On one hand, in comprehensive sex index, outside be being measured except index, still need those who consider webpage of the mass that collects a webpage and different type to distributing. On the other hand, the quality quota system that searchs engine should assure the balanced development of all directions face, the breakthrough that relies on single index can be not improved. Include Yahoo China at present inside the webpage index volume of engine of mainstream Chinese search is in level of 2 billion volume, basically can satisfy the daily inquiry requirement of the user.
However, as a result of from exterior the absolute value volume that cannot measure cipher out to search engine webpage index to measure directly, business of service of all alone engine likes very much search to exaggerate external oneself collecting webpage number, regard the market as stunt. Begin from 1998, krishna Bharat andAndrei BroderBeginResearch, how to watch the volume that index of engine of more different search measures through tripartite guest. After 8 years, this year MayWWW2006On congress, the Ziv Bar-Yossef that comes from Israel and Maxim Gurevich because of this respect outstandingStudy positive resultCarried off the optimal paper award with unique mass rally. Their research cipher out the index of engine of mainstream English search measures opposite volume: Yahoo is 1.28 times of Google, google is 1.36 times of MSN. Are they how cipher out of these numbers? We will all alone for search engine fan introduces this algorithm below, and how is discussing applying on Chinese search engine.
Overview
The index that searchs engine is measured or say to cover the dependency that leads pair of search results, effectiveness for a given period of time gender and find rate have far-reaching effect. Stem from the consideration that the market runs, engine of each big Internet search often announces the documentation measure that he indexes external, however these data often differ level land was joined portion of a few water, there is an unsolved problem on reliability. Accordingly, how to pass those who search engine
Communal interface, the search draw a frame round that says normally namely, the index volume that checks it well and truly became the issue of an attention making a person.
Graph 1, to searching the index sampling of engine
Each index that searchs engine enclothed a subclass of complete documentation on Internet. If we regard pair of this aggregate sampling as the test, so the crucial point of the problem depends on how realizing approximate
Wait for probability random sampling (Uniform Search Engine Url sampler), refer to a graph 1. Specifically, assume S of engine of a search indexed in all | D | Documentation, so we hope sampling gets certain the probability of specific documentation is 1/ |
D | .
Once realized what index through searching casing to be oppositeWait for probability random sampling, we can go up to be sure quite in statistical meaning the opposite volume that the ground estimates to search engine index is measured. Following plan institute show:
Graph 2, search the opposite size that engine indexes quite
We are opposite first N1 of engine S1 random sampling Url. Next, through Url inquiry learned engine S2 to index among them N12 Url, and additional without index N10. In other words, n1 = N10 N12. No less, if we are right N2 of engine S2 random sampling Url, discovery among them N21 is collected by S1 and N20 was not collected, n2=N20 N21. So the opposite volume that we can estimate S1 and S2 is:
| D1 | / | D2|
≌ (N12 N10) / (N12 N12N20/N21)
＝ (N1N21)/(N2N12)
＝ N21/N12 (if N2 of N1 ═ ═)
To be continued. . .
Tags: , Author, engineer, morning, old, States, United, yahoo
