Relevance is an important concept in search engine operation. Search engines must return results most suitable to users for given queries.
There are several different ways and principles for describing relevance. Search engines explore various concepts of relevance. Google recently patented a new search method definition designed to help users get more relevant results.
The patent, in addition to using the search query, introduces the concept of a “label”. A label is an annotation of the query made on pages that may be returned as search results.
If, for example, a user is looking for information about some disease and wants to find its symptoms, the search query might look like:
[disease name label:symptoms]
The result would be search for “disease name”, but pages labelled “symptoms” would rank above the rest.
Personalised search often considers past search and browsing history, but this isn’t always useful — a user doesn’t always want to find the same information as before. Sometimes search engines show suggestions based on actions of users with similar history, but that’s not always needed either.
Labels, however, could genuinely help improve results. But does the search engine create these labels and link them to web pages?
First, many pages don’t contain annotations or labels, even if those pages host useful information. So first we need to solve the problem of how search engines apply annotations to specific pages. The search engine might, for example, consider information about patterns in site page URLs.
The idea of adding “labels” to a query is interesting, but it’s also interesting how many users would use labels in search. Undoubtedly, the ability to use labels would simplify finding relevant pages in specific cases.
Of course, some people could intentionally apply annotations and labels to inappropriate pages or otherwise interfere with label-based search — but most likely such results could be filtered out somehow.
Google already provides the ability to include labels in its custom search system.
