We use search engines to find information about the world around us. In return, search engines are working on using information about how we search and browse web pages to provide us with information that we want to find.
A search engine might gather information from search engine log files as we search to see which terms we use to search with and which pages we select in search results. If we use a search engine tool that may collect information about where we go on the Web, the search engine may also track where we browse when we are looking at pages related to our searches.
How does a search engine use this information about how we search and browse on the Web? What kinds of assumptions are they making about searchers’ behaviors, what searchers might find valuable on websites, and the best ways to get searchers to pages that those searchers might be interested in finding?
A recent patent application from Microsoft gives us some insights and describes search trails, destination pages, interactive hubs, and way stations, all defined below.
Query Suggestions Based upon Searching and Browsing Behavior
When we search, the words we enter into a search box are matched against an index of the Web to identify a good fit with our search terms.
The results we receive include links ordered based upon how relevant those are. The search engine may also provide some query suggestions with those search results.
Some search engines look at how many searchers might change and refine the queries they use during search sessions to come up with query suggestions.
There are some limitations to creating query suggestions based solely upon interactions between searchers and search engines.
If you don’t know much about the topic that you are searching for, and your searches are exploratory in nature, the query terms that you use and the pages that you select in response to those queries may not be helpful to future searchers, especially if you stop searching and start following links from pages that you find to deeper sources of information.
A search engine could look at a mix of search information and browsing information together, to follow a searcher along with their search and browsing sessions, and see what queries a searcher uses, which pages are selected from search results, what links might be followed on those pages, and which pages that searcher finally ends up upon.
We see some of that happening with Google’s site links, which provide shortcuts within search results to the top ranking site for some query terms, to pages within the same domain.
Imagine a system similar to Google’s site links that provide both relevant search results in response to query terms and recommended pages based upon the searching and browsing behavior of people who searched for the same terms. These recommended pages could be from several different domains.
Search Trails: Destinations, Interaction Hubs, and Way Stations
The search engine may find many different types of pages to recommend to searchers.
They define three kinds of pages for us that they might recommend:
Destinations – pages or domains where other searchers have ended up after submitting a query, or more than one query, and end up upon and stopped searching.
Interactions hubs – web pages or domains that other searchers interact with intensively after submitting a query. This kind of interaction involves visitors to those pages viewing pages from those hub pages, and then returning to the hubs to view more pages.
Way stations – pages or domains that others pass through on route to other pages or domains. They might contain little or no relevant information to the query but may be required to travel through to get to pages containing wanted information efficiently.
Links to all three of these types of pages could be provided to searchers as additional information, along with the search results.
The Microsoft search trails patent application is:
Using search trails to provide enhanced search interaction
Invented by Ryen W. White, Mikhail Bilenko, Nicholas E. Craswell, Michael M. Cameron, Hugh E. Williams
Assigned to Microsoft
US Patent Application 20080306937
Published December 11, 2008
Filed June 11, 2007
Abstract
It has been found that user navigation that follows search engine interactions provides implicit endorsement of resources (such as web resources) that are preferred by users and which may be particularly valuable for exploratory search tasks.
Thus, a combination of past searching and browsing user behavior is analyzed to identify additional information that augments search results delivered by a search engine.
The additional information may include a display of hyperlinks to locations derived from the past searching and browsing user behavior, given a specific input query.
The additional information may be provided to supplement web search results on the query interaction logs. Generating a user interface indicates both the search results obtained from the information retrieval and the sites or domains identified separately.
Search Trails and Experts?
One aspect of this patent filing that does concern me is that people might influence the recommendations to show that the search engine might identify as experts. I’m not convinced that the examples of how experts are chosen are experts.
How are these experts chosen?
1) Some experts might be self-identified in their online profiles.
2) Some experts might be identified by how they search at a search engine. For example, if they put phrases in quotation marks, or put a plus sign in front of terms that absolutely have to be returned in search results, or use other advanced search operators, they might be considered to have at least some level of expertise in searching.
3) Some experts might be identified based on how often they look at the information on a particular topic. Someone who frequently submits many queries on “physics,” “astronomy,” and “space” is likely to be more informed about places to look for information about those topics.
While destinations, information hubs, and way stations may be identified by looking at the interactions of all people searching for a specific topic, pages visited by these experts might be given a little more weight. Other methods might also be used to filter pages based upon user expertise as well.
Suggested Sites in Specific Orders and Mis-Typed Queries
When you are learning about a topic, there may be a benefit in visiting sites about that topic in a specific order. Recommendations could help expose a searcher to key concepts for a topic and make it easier to learn about that topic.
This ability might be presented to searchers using a combo box or toolbar plugin or frame launchable from the search page for people who might not have a toolbar installed.
The patent application also describes how this process could be used to provide results for mistyped queries.
Search Trails Conclusion
For Searchers
This might mean to searchers that they might see additional query suggestions along with search results that lead them to destination pages, information hub pages, and way stations.
This means that searchers might be able to find the information they are looking for more quickly with the destination pages, find useful resource pages in the information hubs that are presented, and find way station pages that lead them to invaluable resources that they might not have been able to travel to directly.
For Site Owners
This might mean to site owners that they should consider if their site has pages to consider destination pages or information hubs or way stations.
Are there destination pages on their sites that inform and answer the information needs of visitors well? Do they have information hub pages that people will visit, follow links from, and return to over and over? Do they have way station pages that people use to find answers to the queries they might have or ways to perform tasks on the Web to accomplish?
If not, they should think about how those pages might help visitors to their pages and get visitors to keep coming back and spending time on their pages.
Hmmm.. I hope Google will release a paper or an information like this one. They only tell pageRank and the likes but they didn’t tell us and how they improve or predict our searchers information.
I’d like to see more papers from Google on how they might use information related to people’s searches and browsing sessions, too.
There are a few listed on the Papers written by Googlers page from Google, especially in the section on Information Retrieval.
Hmmm.. I hope Google will release a paper or an information like this one. They only tell pageRank and the likes but they didn’t tell us and how they improve or predict our searchers information.
Hi Arnaud,
We know that Google does do some analysis of where people browse to when they arrive on a web page, and then visit other pages on a site. There’s a good chance that they use an approach like that as part of their analysis of which pages to show as sitelinks when they add sitelinks to search results for pages.
It’s possible that someone at Google has written a paper describing that process as well, though possibly only as an internal document for other people at Google to read. If there is one, I’d love to read it as well.