A misconception about web pages that lingered for a long time on the Web was that most people who visit your web site will enter your site on your home page. Another one is that the meta description you choose for a page will usually be what a search engines shows as the snippet, or summary, for your page in search results.
Search engines have made it a lot easier for a visitor to enter your site at pages other than your home page. And the summary, or description snippet, that those search engines provide about pages listed in search results are more likely to be taken from text on your page that matches the query terms used to find your page, especially if your meta description doesn’t include that text.
You’ve created a web page, carefully chosen a title for that page that carefully describes the contents of that page, and uses a keyword phrase that you hope your audience will use to try to find the page. You created a meta description for the page that is persuasive, engaging, and (you hope) likely to convince visitors to click on the link to your page when they see it in search results.
How likely is it that a search engine will show your page title and your meta description when your page does show up in search results?
The snippet that a search engine might display for a page in search results may come from content on your page instead of from your meta description. Keep that in mind when writing sentences that contain important phrases that your page might be found for in search results.
Yahoo Research on Page Titles and Search Snippets
Tapas Kanungo was one of four researchers from Yahoo last year who published a paper describing how that search engine might choose what to display as a page title in search results, in Web Search Result Summarization: Title Selection Algorithms and User Satisfaction. In that paper, we are told that there are a number of potential sources that all of the major search engines might consider as candidate choices for use as the title display for a page, including:
- HTML title,
- Anchor text,
- Internal anchor text,
- Open directory page title,
- Various HTML headline titles on the page,
- Yahoo! directory page title,
The paper was interesting for a couple of reasons. One of them is its description of how a search engine might choose a title to display for pages in search results. The second is that it notes that Tapas Kanungo is now at Microsoft, which seems to be a future home for other Yahoo search engineers. His home page at Kanungo.com tells us that he is now a Principal Applied Scientist at Microsoft’s Bing R&D Division. It also tells us that he proposed and was the tech lead for the machine learning process used to create summaries for search results at Yahoo, serving approximately 3 billion queries a month.
Search engines will sometimes use the meta description that a publisher has created for a web page as the summary used to describe the page, especially if it contains the query terms used to find that page. But, it’s just as likely that a search engine will decide to create a summary based upon content found within the page instead. Especially if the words used in the query the page showed up for in search results aren’t included in the meta description.
While a meta description describes the content of a page, a snippet or summary in search results for a page often tries to let searchers know why a page is being shown to them, by displaying text from a page that contains the keywords they used to search for that page.
In addition to the white paper on title selection for search results, Tapas Kanungo wrote some other papers that provide some more insights into processes that may have been involved in creating a machine learning summarization process:
- Predicting the Readability of Short Web Summaries
- Machine Learned Sentence Selection Strategies for Query-Biased Summarization
He is also listed as an inventor on a couple of Yahoo patent applications that I wrote about in the past, in my posts:
Those posts give us some hints as to why a search engine might choose the sentences that they do when deciding upon a snippet to display in search results. Another Yahoo patent application published this week which was co-authored by Tapas Kanungo adds an important detail in that selection, describing something called a “utility metric,” which measures a number of different factors. The patent application is:
Generation of Search Result Abstracts
Invented by Tapas Kanungo, Jan Pedersen, and Tamas Sarlos
Assigned to Yahoo! Inc
US Patent Application 20100057710
Published March 4, 2010
Filed: August 28, 2008
The utility metric that might be used to determine which sentence or sentences, or parts of sentences, might be displayed as a summary of a page in search results may look at factors such as:
- How readable the phrase is,
- How relevant the phrase is to the query,
- How relevant the phrase is to the page it appears upon,
- How long the phrase is,
- Combinations of the above and/or
- Similar considerations.
Conclusions and Search Snippet Takeaways
My takeaways, after reading this patent filing, and the previous patents and papers on search snippets?
1. Try to write a meta description for your page that does a good job of describing the content of the page, includes the most important keyword phrase for that page, and that might persuade people to visit the page if they see it in search results.
2. Anticipate that other phrases on the page might also be shown in search results, and include those in sentences that are well written, and that might be engaging and persuasive if seen out of the context of the page itself. If they match up with a query the page is found for in search results, and are chosen as snippets, or summaries for the page, you want to make it more likely than not that people will visit the page if it is relevant for the tasks or information needs that they are trying to meet.
3. How relevant text within the content of a page might be to a query, or to the page it appears upon depends upon a number of factors that I described in my previous post on How a Search Engine May Choose Search Snippets, in the sections on query dependent and query independent relevance.
How likely is it that Microsoft’s Bing might start considering factors like these in the future? A better question might be why they wouldn’t be considering them now. I think regardless of how many Yahoo researchers and technicians move to Bing, ideas about what make good summaries for pages shown in search results are likely things explored by all search engines.
On the migration of researchers from Bing to Yahoo, Jan Pedersen, also listed as an inventor on this patent, was a researcher at AltaVista and joined Yahoo when Yahoo acquired that company. His home page states that he is now “currently Chief Scientist for Core Search at Microsoft.”