Method for estimating coverage of web search engines

1. A computerized method for estimating coverage of search engines, each search engine maintaining an index of words of pages located at specific address in a network, comprising the steps of:

generating a random query, the random query being a logical combination of words found in a training set of the pages;

submitting the random query to a first search engine;

receiving a set of URLs in response to the random search query;

randomly selecting a particular URL identifying a sample page;

generating a strong query for the sample page;

submitting the strong query to a second search engine;

comparing result information received in response to the strong query to determine if the second search engine has indexed the sample page; and

generating an estimate of the relative sizes and amount of overlap of the indices of the first and second search engines.

2. The method of claim 1 comprising estimating relative sizes of the indices of the first and second search engines by dividing a fraction of a first set of pages sampled from the second search engine that are contained in the first search engine by a fraction of a second set of pages sampled from the first search engine that are contained in the second search engine.

3. The method of claim 2 comprising estimating a relative amount of overlap of the indices of the first and second search engines by computing a fraction of a set of pages sampled from the second search engine that are combined in the first search engine.

4. The method of claim 1 wherein the training set of pages relates to a particular context domain.

5. The method of claim 1 wherein the random query combines random words selected from the training set with a logical operator.

6. The method of claim 1 wherein the random query is a disjunctive query.

7. The method of claim 6 wherein the disjunctive query combines a set of words using OR operators, the set of words having a predetermined size.

8. The method of claim 7 wherein the words of the training set have relative frequencies that are substantially similar.

9. The method of claim 1 wherein the random query is a conjunctive query combining a pair of words and an AND operator.

10. The method of claim 9 comprising:

sorting the words in the training set according to frequencies of the words; and

establishing an upper frequency threshold and a lower frequency threshold so that when words equidistant from the upper and lower thresholds are combined in the conjunctive query, a set of addresses is less than or equal to a predetermined maximum number of members.

11. The method of claim 1 wherein the network is the World Wide Web and comprising:

fetching the particular URL from the first search engine;

fetching a corresponding page from the World Wide Web; and

constructing the strong query to be representative of the sample page.

12. The method of claim 1 wherein the result information includes URLs of pages indexed by the second search engine.

13. The method of claim 12 wherein the URLs of the pages indexed and addresses identifying the sample pages are normalized before the comparing.

14. The method of claim 12 wherein the result information being compared is content of the sample page, and content of the pages indexed by the second search engine.

15. The method of claim 12 wherein the result information includes host names.

16. The method of claim 1 comprising discarding dynamic and outdated pages before comparing.

17. The method of claim 1 wherein privileged access is provided to the first search engine.