Other advancements include identifying key moments within video content, as well as further enhancements to BERT language understanding and interpretations of miss-spellings.
Why is passage ranking useful?
By ranking passages in their own right, Google can improve the quality of answers to specific searches, making it quicker and easier for the user to locate the information they are looking for.
Whilst many searches are satisfied by the most relevant websites or webpages on the topic, some queries are far more specific. The exact information you are searching for may not be immediately obvious from a page on the topic. In fact, it may actually be best answered by a passage within a page on a different topic.
As Google puts it, “sometimes the single sentence that answers your question might be buried deep in a web page.” This new functionality means they can “find that needle-in-a-haystack information” that users are searching for.
puts it, “sometimes the single sentence that answers your question might be buried deep in a web page.” This new functionality means they can “find that needle-in-a-haystack information” that users are searching for.
How does it work?
As always, Google Search’s primary aim is to match the most relevant information to the search query.
Previous Search AI relied heavily on page titles and heading tags to rank a webpage against the search query. But for very specific questions, a webpage that ranks highly on the keywords in the search bar may not give a direct answer to the actual question. The new AI is able to harness content from a far wider range of webpages – such as forums – to pinpoint the exact answer to the question provided.
For example, say you were wondering how to tell if you windows are UV glass. Google might previously have directed you to a page such as the one on the left in the image below. This page covers the topic of UV radiation through windows and therefore would hit all the important keywords.
However, as you see on the right, a far more direct answer comes from a passage within a forum on a DIY site. In the second result, the webpage may not have been as highly connected to the search query, but this particular passage gives the most direct and helpful answer.
How is this different from Featured snippets?
Google is already able to identify relevant information within the body of webpage text and feature selected paragraphs on the results page. So how is passage ranking different?
The main change is that the passages themselves are now used to determine the ranking. With featured snippets, the ranking criteria are based on the page as a whole – focussing on signals such as page title and headings - and then the most relevant passage from each page is subsequently featured.
As a result, the featured snippet, while being the most relevant passage within the ranked page, may not be as helpful as a passage from a page that ranks less well as a whole.
Is this the same as passage indexing?
Another key difference to note is the distinction between passage ranking and passage indexing. Whilst Google’s announcement makes reference to indexing individual passages, the change is in reality an issue of page ranking, not indexing. Google still indexes whole webpages, as before. However, one can assume that the content in all passages on the page is considered and will inform the indexing going forward.
How will this affect SEO?
From an SEO perspective, being able to rank for individual passages within an article, as well as for the article as a whole could open up many more opportunities to rank.
While Google’s Martin Splitt has indicated that there is no immediate action required from content writers in response to the new algorithm, the change is still good news for long-form content pieces.
If you are writing on a broad topic covering a range of subtopics, you will now how the chance to rank for specific subtopics in their own right. Passages within the body of your content may start ranking of their own accord with non-main keywords.
Much of this new functionality in Google AI is thanks to its improvements in natural language processing over the past years. Google’s BERT algorithm is increasingly capable of interpreting nuances in language and predicting meaning in its own right. This enables it to be far more agile in the way it trawls webpage content, relying less on key indicators and more on an intelligent understanding of meaning within language.
There can be no doubt that this will be just one of many more similar advancements to come.