Google Multitask Unified Model (MUM) is one of the biggest updates the search engine giant is developing. It was introduced in May 2021 during the Google IO Conference. This is perhaps the biggest of all Google updates 2021, and it’s already improving search results today.
MUM is comparable to its predecessor, Google BERT, since it too is capable of natural language processing (NLP). Authorities from the tech giant are saying the update is “1,000 times better” than BERT AI.
In this blog, we’ll cover:
• What is NLP?
• MUM meaning
• The Google BERT update
• Google NLP AI
We’ll also talk about MUM’s impact on search engine optimization (SEO).
Why Launch Another Google Search Update?
There will always be a need to update Google Search. Search engines started with keywords, followed that with key phrases and now we’re entering the realm of natural language processing through the Google BERT update and now MUM.
According to Google Search Vice President Pandu Nayak, the goal behind MUM is to improve user experience (UX).
BERT, Google’s original contextual learning technology, was deployed back in 2019. It improved the search engine’s ability to figure out the context of human speech. MUM, Google’s latest update, builds on what its predecessor has started. The Multitask Unified Model goes beyond just context. With this latest Google Search update, it’s going to feel a lot more like you’re talking to an actual person.
The Quest for Understanding Human Speech Better
Let’s say you want to know the most authentic way to cook adobo, a local Filipino dish. People would use several search phrases, such as:
• “How to cook adobo”
• “Authentic adobo recipe”
• “Best adobo recipe”
• “Cook adobo like a Filipino local”
According to Nayak, people use eight search phrases on average to get answers to similar complex questions. Researching takes a lot of time, and there are several nuances to complicated questions.
BERT, Google’s first NLP technology, can determine search intent. It’s capable of natural language understanding. It can determine that you’re looking for Filipino adobo – not the Mexican marinade or the Portuguese stew of the same name.
Like Asking an Actual Person
Google MUM takes things up a notch. With this search update, Google’s search engine will be able to understand the various nuances in linguistics and speech. This is much more advanced than the last Google BERT update.
You don’t have to use eight search terms – you only need one. It’s like asking a Filipino chef how to cook authentic adobo. She will walk you through the entire process without the need to ask multiple questions.
Dealing With Language Barriers
This is another reason why Google MUM AI is a big improvement on BERT: It can understand other languages. The Multitask Unified Model update is trained across 75 languages.
“Training” in Google NLP terms means different language data has been fed to MUM’s machine learning AI. (We’ll elaborate further when we tackle the question of what is natural language processing in AI.) This feature gives the Google MUM AI the ability to take information from websites written in different languages other than English. It can use foreign language websites to answer your questions.
What Is Natural Language Processing in AI?
What is NLP? Let’s begin with a beginner-level NLP definition: NLP is actually a collective term that is used in different disciplines, such as the fields of artificial intelligence, computer science and linguistics.
In simple terms, NLP is all about helping computers achieve natural language understanding. It goes beyond providing definitions for each word. The functions of the words and phrases are also processed in order to derive meaning and context.
To help make sense of our NLP definition, we need to understand how it works. In natural language processing, human language is broken down into smaller logical segments. This includes words, phrases and the different parts of speech. Grammar rules are also applied.
Breaking Down Human Speech
For a deeper understanding of NLP, it helps to know how Google BERT and BERT AI work. Here’s a quick walkthrough of the basics of natural language processing and how an NLP API processes human speech:
1. The search phrase that you enter into the Google search box is broken down into smaller semantic units (e.g., words, phrases and clauses).
2. The words in each semantic unit are then tagged according to their part of speech, such as verbs, nouns, adverbs, adjectives etc.
3. Words are then reduced to their root forms, which standardizes them.
4. After that, words that add little to no information are removed.
5. Algorithms (e.g., BERT and MUM) are then applied to the data that has been collected. Some algorithms make use of a rules-based approach, while others apply statistical methods. MUM uses the latter.
The Google BERT NLP API goes beyond just the meaning of the words that can be found in a search query. It also analyzes the context of the entire search phrase and the relationships of all the words in that phrase or sentence.
BERT stands for Bidirectional Encoder Representations from Transformers. It analyzes the words that go before and after the search terms; ergo, it is bidirectional. The relationship of these words is established, thus allowing BERT to better understand the context of an entire sentence.
What about the term “transformer?” In machine learning, transformers refer to the part of an algorithm that can change the sequence of the words. This gives the Google AI the ability to know which words are more important. It can rearrange the sequence of the words without losing the actual meaning of sentences.
The newer Google AI used by MUM has that contextual ability, combined with multitasking and multi-modal capabilities. It also derives information from images and other data types.
Real-World Performance
How does Google Multitask Unified Model fare in the real world? And how does it compare to BERT?
When applied to a question answering system about COVID-19, BERT produced information with 58.1 percent accuracy. The interesting part is that the word “COVID” wasn’t used in the search, yet the results were very accurate. BERT is effective at accounting for the context of the search terms.
Other research shows that BERT is very good at predicting missing words. It can suggest search terms and keywords that might be helpful to you as you search for information.
Google MUM, on the other hand, can do what BERT does when it comes to contextual analysis. Even if it’s still in beta phase, it curates information efficiently from different sources and in different languages. You can say that it transforms your search into research instead.
During Google I/O, Prabhakar Raghavan, a Senior Vice President at Google, demonstrated MUM’s capabilities, using a full-sentence search query: “I’ve hiked Mt. Adams and now want to hike Mt. Fuji next fall, what should I do differently to prepare?”
In this simulation MUM was able to display similarities and differences between the two mountains and analyze photos of essential mountain climbing gear that may be useful. This demonstrates MUM’s multi-modal and multitasking capabilities.
In another Google blog post, Pandu Nayak explained how MUM improved searches for vaccine information. MUM was able to identify more than 800 variations of vaccine names from 50+ languages – and it was able to get all that information in a matter of seconds. For comparison, this type of task usually takes several weeks to perform if you do it using the old method of multiple search terms.
This suggests that MUM doesn’t only identify related search terms or missing words like BERT; it can also pull information from different languages and identify the different ways people refer to COVID vaccines quickly and efficiently.
What Does This Mean for Your Business?
Google MUM is bound to have a huge impact on your online presence. If Google is successful in this endeavor, it will create a cohesive search methodology. This, in turn, will change search behavior – and that will impact how you do SEO.
There are some who believe that MUM will kill SEO, but John Mueller, Google’s Senior Webmaster Trends Analyst, says this won’t be the case. It shouldn’t make SEO obsolete – but it still has the power to change search as we know it.
As with most Google updates 2021, this will require more focus on technical SEO to improve your site’s performance. You will still need to do site audits, improve your content, perform keyword research and earn backlinks. But you need to prepare for MUM’s impact on your SEO efforts.
Partnering with a full-service digital marketing agency like Thrive will help you prepare your digital marketing strategy for Google MUM and any other future Google Search update. Collaborating with search engine marketing experts will ensure that your business ranks at the top of search results no matter what update Google Search introduces next.
Call 866-434-4748 today to schedule a free consultation or fill out this form.