Last updated on
Google’s recent Search Off the Record provided a deep dive into the inner workings of Google Search. During the event, John Mueller and Lizzi Sassman from Google engaged in a discussion with Elizabeth Tucker, Director of Product Management at Google. She shed light on the intricate systems that collaborate to determine the ranking of web pages, highlighting the importance of systems such as topicality in this process.
The term “topicality” typically refers to how relevant something is at the present time. However, in the context of search, “topicality” specifically refers to matching the subject of a search query with the content available on a webpage. Machine learning models, such as BERT (Bidirectional Encoder Representations from Transformers), play a crucial role in helping Google interpret user intent by understanding words in relation to their surrounding context.
Elizabeth Tucker highlighted BERT as an example during the discussion, emphasizing its ability to grasp the meaning of a word within the context of nearby words. She underscored the importance of aligning content with the topical relevance of a search query to enhance user satisfaction.
During the conversation, Lizzi Sassman inquired about user satisfaction, prompting Tucker to explain that search encompasses multiple dimensions and systems. She illustrated this complexity by stressing the significance of ensuring content matches the topical relevance sought by users.
During the discussion around the 4:20 minute mark, Lizzi posed a question:
“When you mentioned satisfaction, are there more detailed ways we’re examining? What constitutes satisfaction after conducting a search?”
Elizabeth responded:
“Absolutely, Lizzi. Within Search Quality, we consider numerous critical aspects of search. We operate several systems aimed at presenting content that aligns closely with the search topic. In the early days of Google Search, this posed challenges at times.
While our systems have significantly improved, there are still instances, particularly with complex queries, where we may encounter difficulties. People search in diverse ways—from entering keywords to conversing with Google using everyday language. I’ve witnessed remarkable queries like, ‘Hey Google, who was that person years ago who did that thing I can’t recall?’ These lengthy, vague queries are now being answered by our advanced systems.”
Key insight:
A key takeaway from this exchange is the intricate interplay of multiple systems, where topical relevance is just one component. Often, the search marketing community emphasizes singular factors like Authority or Helpfulness, yet in reality, search involves numerous dimensions. Oversimplifying these factors to one, two, or three concepts can be counterproductive.
During their conversation, John Mueller inquired with Elizabeth about biases in search, and whether Google considers this aspect. Elizabeth acknowledged that Google actively monitors various biases and strives to mitigate them. She explained the nuances of different types of search results that may be relevant, such as evergreen content versus fresh updates, emphasizing the delicate balance Google aims to achieve.
John’s question, around the 05:24 minute mark, was:
“When analyzing data, I assume biases come into play. Is that something we also take into account?”
Elizabeth responded:
“Absolutely. There are numerous biases we’re mindful of when users seek information. Are we disproportionately favoring certain types of websites? Are we prioritizing encyclopedic or evergreen content over current updates? Are we showcasing content from major institutions or smaller blogs? Are we highlighting voices from social media platforms?
Our goal is to ensure a diverse mix where we can surface high-quality web content in all its forms and sizes—modest goals, indeed.”
This exchange underscores Google’s commitment to addressing biases and maintaining a balanced approach in delivering search results.
Elizabeth emphasized her involvement with a diverse array of search systems, highlighting a critical point often overlooked by the search community: there are numerous systems beyond the few commonly known ones.
It’s essential not to narrow troubleshooting efforts solely to one, two, or three systems when addressing ranking issues. Instead, maintaining an open perspective is crucial, considering that the root cause could be something entirely different, unrelated to factors like helpfulness or EEAT (Expertise, Authoritativeness, Trustworthiness).
John Mueller questioned whether Google demotes websites based on user complaints about specific search results.
Elizabeth clarified that the majority of the systems she oversees do not involve demoting sites. She underscored the complexity by noting her work with multiple systems and signals, extending beyond the limited set typically emphasized by the search marketing community.
Among these systems, Elizabeth referenced the core topicality systems. This term suggests the existence of several systems and algorithms aimed at aligning search queries with relevant topics. She explained that these “core topicality systems” are designed to ensure accurate topic matching for search queries.
At the 11:20 minute mark, John inquired:
“When there’s significant public outcry, is the initial response to demote a site that’s clearly problematic, or is there a balance between promoting positive content and demoting less desirable results?”
Elizabeth responded:
“That’s a great question. My role in Search Quality involves managing various systems, which is an exciting aspect of my job. We rely on numerous signals and systems working together to deliver high-quality search result pages.
Some systems inherently focus on demotion, such as those addressing web spam. For instance, if we detect malicious download sites, our aim would be to identify and mitigate these issues to safeguard users.
However, the majority of systems I oversee are geared towards promoting relevant and valuable content. For example, I’m involved in developing our core topicality systems, which aim to align search queries with appropriate content.
Matching keywords is straightforward, but understanding natural language remains a complex challenge. Recent advancements in machine learning have greatly improved our ability to grasp user intent and deliver relevant content. This is an ongoing, intricate task.
In exploring topics like topicality—admittedly a bit of a technical term—we’ve discovered that improving this aspect enhances our capability to handle more nuanced and challenging search queries.”
This exchange underscores the complexity of managing search quality and the emphasis on promoting relevant content while addressing problematic sites.
Elizabeth revisited the concept of Topicality, now referring to it as the “topicality space,” and emphasized Google’s extensive efforts to refine this aspect. She contrasted Google’s historical focus on keywords, noting a significant shift towards prioritizing topical relevance.
She discussed this at the 13:16 minute mark:
“Google used to heavily prioritize keywords. Simply stringing together words with prepositions could lead to inaccuracies in search results. Prepositions posed significant challenges for our systems, particularly in earlier times. Looking back, it’s somewhat humorous to reflect on these challenges.
In the past, searches typically consisted of one, two, or three keywords. When I joined Google, anything beyond four words was considered lengthy. Nowadays, it’s common to encounter searches spanning 10 to 20 words or more. Managing these longer queries and identifying the key words within them presents new challenges.
An example from years ago illustrates our evolution: queries phrased as questions, such as ‘how tall is Barack Obama?’ These queries required pages that provided direct answers, not merely matches for the words ‘how tall.’ The inception of featured snippets was a response to this need—to match answers, not just keywords in questions.
Over the years, we’ve dedicated substantial effort to what we term the ‘topicality space.’ This remains an ongoing focus for us.”
Tucker’s response underscores several key insights, particularly the importance of considering Google’s core topicality systems when analyzing search ranking algorithms. These systems help Google understand the topics of search queries and match them with relevant webpage content. This highlights the shift towards prioritizing topics over solely focusing on keyword optimization.
A common error I often observe among those struggling with rankings is an excessive fixation on keywords. Over the years, I’ve advocated for an alternative approach that emphasizes thinking in terms of topics. Unlike the one-dimensional approach of optimizing for keywords, focusing on topics is multidimensional. It aligns closely with how Google ranks web pages, where topical relevance plays a crucial role in ranking decisions.
Original news from SearchEngineJournal
