Saturday, February 28, 2009

Snippets: Another key to searching

While federated searching brings together a broader field of information for single-point searching, there's nothing especially intelligent about the results. The idea behind federated or meta searching is that it makes the Deep Web more accessible, by enabling searchers to tap into numerous databases at once. The Deep Web is only deep because the information you need is not in the database where you are searching. Knowing WHERE to look for information is critical and federated searching makes that easier (although there is no engine I know of that looks everywhere).

By looking in more databases, your chances of finding what you are looking for ought to improve. But unless you are new to searching and think the top results are always the best ones, pulling results from 10 databases is really no better than pulling them from one. One reason is the keywords you start with. You typically have to speculate about the words to use. Most of the time there are better words that could be used, but you don't know them yet.

To a limited extent, federated searching takes care of speculating where to look for information. But it doesn't eliminate the need for a critical skill you need once you obtain results.

Checking the relevance of results obtained remains extremely important. Effective searchers check how their keywords are used in the snippets, or abstracts, that are returned. If you know what you were looking for, the language of the snippet usually makes it clear whether the information is relevant--whether it makes sense.

The other thing scanning snippets provides is better keywords. Since the words you start with are not always the words you actually need, being observant for better words pays off in terms of search efficiency. A classic example is searching for 'buffalo', a term with many meanings. If you are looking for the number of buffalo alive today in North America, the term 'bison' actually shows up in the snippets. Bison is a better keyword that improves the initial query, because it has only one meaning.

Here are a couple of queries to try. It doesn't matter much which search engine you use (I used Google). Look for relevant results and better keywords in the top ten snippets. If you want to, submit the answers you find (not just the ones you think of) by commenting to this blog.

Relevancy Challenge: Other than the fastest time from bottom to top, what is another record for climbing the Sears Tower?
Suggested query: sears tower climb record
Scan the snippets for an answer other than the fastest time.

Better Keyword Challenge: What types of shoes have been used for climbing the Sears Tower?
Suggested query: sears tower shoe climb
Scan the top ten snippets for names and types of shoes. How many can you find? Are all your findings relevant to the question?

No comments: