Understanding Google’s MUM update
Reading time: 5mins
MUM is an AI-powered algorithm that aims to understand semantics and natural language not just within text, but audio, video, and imagery to better aid the user in their search.
It can translate 75 languages and analyse a much wider set of knowledge that can aid in delivering results for complex queries.
What is Google’s MUM update?
Google’s MUM update stands for multitask unified model which was announced in May 2021 at Google’s developer conference I/O. As with all Google updates, it aims to improve search results, this time, using an AI-powered algorithm that will enable it to better understand language and content from a variety of sources and formats.
If we refer to Google’s mission, they want “to organise the world’s information and make it universally accessible and useful.” Even though we are still in the early days of MUM, this new update aims to deliver on just that with a few innovative and ground-breaking changes.
When we head to search, more often than not, we do not always know how to craft the best query to return the result we desire, as we may not have the knowledge to begin with, we head to Google to learn.
Over the years Google has made steps to improve their understanding of Semantics (semantics is the study of the meaning in language, it can be applied to a single word or whole paragraphs) and how we as humans interact with search engines, the time has passed since we once put just keywords into Google and now engage in a conversational way.
MUM and previous updates utilise the latest machine learning science to improve and help Google understand these queries no matter how you spell or combine words.
MUM is based on semantics and natural language understanding, but it is far more advanced and powerful than just that. Its goal is to connect the user with complex answers to their queries regardless of language or media formats. MUM is 1,000 times more powerful than the BERT update, and also uses the T5 text-to-text machine learning framework.
Google states that it takes, on average, 8 searches to get the user the best result for complex queries. MUM can help bridge the gap, not only can it extract data and information from many formats (image, video, audio), but it is also trained in 75 languages, meaning it can analyse and understand a much wider set of knowledge that can aid in delivering results for those complex queries.
It is important to note that previous updates have built the foundation for what MUM is today, all based around semantics and understanding how humans interact with search engines, below are the two most important to note, BERT and Hummingbird.
A brief history on BERT and Hummingbird
In 2019, Google announced what they considered the most important update in 5 years, Bidirectional Encoder Representations from Transformers (BERT) which began rolling out on October 25th. This update focused on understanding complicated search queries that rely on context.
By improving natural language understanding (NLP), particularly for conversational/longer queries, BERT can better understand the context of words within searches to better aid the user find the result they need.
As humans, we can understand certain meanings and nuances in words without the need for explanation, BERT improves understanding for these types of queries, especially for those that include “for” and “to”, which matter a lot for context. As BERT aims to understand the nuances in queries by looking for context in words before and after other words, it will impact both rankings and featured snippets.
One limitation that Google experienced with BERT was that it pushed its hardware to the absolute limits, meaning that it could only process it on 1 in 10 search queries in the US only.
BERT is also open-source, with others having created variations of this technology.
Hummingbird was announced by Google back in 2013. The update was quite a shift for Google’s internal technology as it resulted in a complete overhaul of the core algorithm, whereas previous updates like Panda and Penguin were bolt-ons.
Hummingbird was Google’s shift to the world in that they were attempting to understand queries at a much deeper level by understanding the intent which would match more relevant results to the user. The update aimed to take a broader context into account and not rely solely on matching individual keywords, really this is Google’s first step in building a semantic web.
You can think of Hummingbird as paving the way for mastery in conversational search and the foundations for voice, BERT, and MUM.
MUM, Rich Media and Language Expansion
In 2022, searches are no longer defined by simply typing text into a search bar, we can leverage the power of smartphones to search with many media formats.
MUM can not only understand text, but image, video, and audio files to return results to queries. This can make the search engine result page (SERP) much more diverse, so just relying on text-based content will become a thing of the past.
MUM is also a fantastic linguist, being able to read, translate and understand 75 different languages, so far. MUM can transform knowledge across languages, say so for example an article written in Polish is relevant to your query, but you didn’t conduct the search in Polish, MUM can transfer and leverage this knowledge to return better results in your own language results page, or even translate this page.
What does this mean for you & your optimisation?
What can we, SEOs, webmasters, businesses, and website owners do to be in MUM’s good books? Simple, continue as you were before, assuming you were creating a great user experience and content.
Here are a few tips and ways we can explore:
- Create high quality content that always focuses on the end user, yet opens up for contextual relevant internal and external linking for related topics
- Increased emphasis on multimedia elements such as high quality imagery, videos and podcasts. If you have social media channels and third-party hosting platforms, optimise them fully and make sure they link back to your website
- Continue to build brand awareness and loyalty with consistent use of brand voice across all touch points
- Add structured data to give search engines clues about your on-page content
- With the breakdown of languages and potential increased competition, continue to build your expertise and authority in your niche or industry
Final words from our SEO team
What is important to remember is that whenever Google releases a new update that will impact Search, it goes through a rigorous screening and testing process to make sure more relevant and helpful results are being returned. That being said, MUM is based on a constantly learning, AI-powered algorithm, they may not get it right 100% of the time straight away, but Google believes it will evolve and deliver exactly what they want, a more unified, universal, and accessible Search.
This update will be rolled out constantly over the coming years, with every iteration carefully monitored to make sure no bias or one-sidedness.
Get in touch to speak to an expert to learn more.