Information Technology Reference
In-Depth Information
“memory” separately and then find which pages contain both refer
ences. Presumably the pages you obtain would include references
for “virtual memory,” but other references might be included as
well. For example, if one page indicated that “virtual reality appli
cations require a significant amount of main memory,” then that
page would be retrieved as having references to both terms. On the
other hand, if you conduct your search with double quotes around
these two terms, then the search engine will look for only those
pages where the word “memory” comes immediately after the word
“virtual.”
Just as with searching for material in a book, search engines pre
pare their own indexes for finding materials related to your queries.
As with the storage of Web pages, search engine companies utilize
two approaches for the actual preparation of these indexes: human
and automated compilers. In their early years, both Yahoo! and Ask
Jeeves employed people to scan pages, select the most helpful pages
for their indexes, and identify themes for each page. Software tools
developed by the vendor then would compile lists of themes and
pages from the indexers and provide the search engine with the in
formation necessary to respond to queries. An important advertis
ing point of each of these sites was the human oversight to their
search results; sites selected by people were thought to be generally
more helpful than those selected by machines. Thus, even though
queries to search engines with humancompiled indexes might result
in relatively few responses, the sites that were identified often were
of fairly good quality.
The second way to compile indexes is to turn the problem over
to programs. In this approach, the challenge is to clarify relevant
rules for assessing the content of a site so that an index contains
only relevant references. Due to practical constraints, current search
engines often maintain several indexes to address this issue. For ex
ample, one index might consider keywords specified by the author,
another index might capture the title of the document, another
might look at the body of the page, and another might scan for
words in the URL itself.
However indexing is done, indexes must be updated whenever
new pages are scanned, and index entries must be revised when new
copies of pages take the place of earlier versions.
