03.01.2015 Views

Combining Information from Multiple Internet Sources

Combining Information from Multiple Internet Sources

Combining Information from Multiple Internet Sources

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

3.2 Auction method<br />

Auction method as its name tells is an auction adjusted to be used in the thesis. Unlike real<br />

life auction the one which is implemented consists only of buyers. They try to reach an agreement<br />

on price of the commodity (URL) – select the product with lowest price.<br />

This approach was also used before in the NeurAge system [9], and has been adapted to be<br />

usable for purpose of this thesis. Like in the algorithm described previously, the Auction method in<br />

its original form was about agents voting about the classes of data. In this adaptation classes were<br />

replaced by URLs and agents are supposed to vote about those. This method also returns 10 distinct<br />

URLs like the Game theory method with the same assumptions concerning correctness.<br />

In each round of the auction each agent has its product (URL) assigned. Afterwards, the<br />

“cost” for each assigned URL is calculated. Costs are compared and the agent with the highest cost<br />

is considered to be a loser. Afterwards, the confidence values for selected URLs are updated by<br />

subtracting the cost <strong>from</strong> their value. Henceforth, the next round takes place. If the agent that was<br />

marked before as a loser loses again, it and its result set are discarded <strong>from</strong> further negotiation<br />

process - each agent has two chances before being removed. After removal, the process enters its<br />

next round, and so on until one agent remains with his selected answer. This is repeated 10 times<br />

and therefore it presents the same as the Game theory based approach way of evaluation of its final<br />

answer. As in the Game theory algorithm; each time the URL that was selected already as a part of<br />

the final answer, is not included in next “big” rounds of the Auction process.<br />

In this algorithm (as in the previous one – the Game theory) the URL ranking is used as a<br />

base for all calculations. Therefore at the beginning of the main process we apply the algorithm<br />

<strong>from</strong> section 3.4.1. The main process may start or may not, depending on outcome of this algorithm.<br />

Auction method also requires that search engine has opinion on every URL that takes part in the<br />

negotiation process, just like the Game theory approach. The result sets which do not contain a<br />

particular URL are updated with the aforementioned URL. If the results sets are completely disjoint<br />

the combined result set created <strong>from</strong> all result sets without URLs repetition will be returned. In this<br />

case feedback to the application should be provided so that the search engines can be ranked. If it is<br />

not the situation, then the main part of the Auction starts and afterwards 10 results will be yielded at<br />

its conclusion.<br />

Listings 3.2.1 and 3.2.2 present the example of flow of the one round which is a part of<br />

Auction method. All calculations that take place in the round of the Auction are presented in this<br />

example; on real numbers. This example also shows how the algorithm behaves in particular<br />

situations.<br />

28

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!