Unranked Results: Not as fun as ranked results …

It’s self-evident that search engines return results. What many people don’t realize, is that search engines return two basic types of results: rank results and relevant results. Ranked results are preferred, because this means the search engine used some algorithm to determine the relative importance of each result from one another and (hopefully) returned a sorted list with the most relevant result on top. Google is an example of such a search engine, as is our search engines, including WorldWideScience.org, Science.gov, Scitopia.org and mednar.

However, many search engines return merely relevant results, and don’t expend any effort in attempting to rank the results. This means the list of results are probably relevant, but you have no idea which item in the list is the most relevant or important to your research. In other words, page 20 of your results list could contain the proverbial needle-in-the-haystack, but you’d never find it, unless you click and evaluate each and every article in the results list from page 1 to page 19, before your find what you really wanted. PubMed is an example of such a search engine, and it happens to be very important and popular in the medical community.

In federated search, you get both: Ranked results and relevant (but unranked) results. The reason for this, is because our federated search technology uses the meta data returned from each search engine queried to establish ranking. Check out my previous article, Ranking: The Secret Sauce for Searching the Deep Web, on how we specifically conduct our ranking. The problem is, the meta data returned from each search engine federated may not contain the search terms that were used in the search. When this happens, our federated search technology has a result, but is unable to establish its rank in relation to the other results. A result becomes an “unranked result,” in that it’s relevant but its rank is unknown.

Perhaps it’s best to give an example: Suppose you search for “blue laser semiconductor.” A result that contains only “blue laser semiconductor” in the title will be ranked very high. However, a similar article entitled “high-wavelength electronic media” will be unranked, unless “blue”, “laser”, and/or “semiconductor” appear in other metadata, such as the abstract. We know the article is relevant, because it was referenced by a search engine, but we don’t know how relevant because we’re not able to find the search term(s) in any of the metadata returned by the search engine being federated.

When you perform a search using our federated search technology, any results that are returned by the search engine which does not contain any of the search terms, or doesn’t contain the exact phrase searched for, in the title or abstract, are considered unranked by Explorit.

There is no particular order that unranked results are displayed, except for perhaps in the order that they arrived in, as they all have a ranking score of 0.

In some cases, bad results are returned by collections because they have a low-quality search engine that doesn’t handle multiple Boolean terms, exact-phrase searches, or large-phrase searches, well or at all.

The more complex the query you issue, the fewer number of ranked results you will get. Our recommendation, when using federated search, is to try simpler queries first.