Over 10 years we helping companies reach their financial and branding goals. Onum is a values-driven SEO agency dedicated.

CONTACTS
News SEO

What is Google MUM and Why Should You Care?

what is google mum - mount fuji

What is Google MUM and Why Should You Care? – In this article I will explain what is Google MUM.

You will learn what is Google MUM, why it’s important for search and how it compares to BERT.

If you need help with increasing the visibility of your website on search engines, make sure you check our SEO services.

What is Google MUM?

MUM stands for Multitask Unified Model and is a new AI-powered technology focused on understanding and answering complex search queries on Google.

https://www.youtube.com/watch?v=s7t4lLgINyo

What is an Example of a Complex Search Query?

The example given to show how MUM works if the following:

I’ve hiked Mt. Adams. Now I want to hike Mt. Fuji next fall, and I want to know what to do differently to prepare

How Does Google Treat Complex Search Queries Today?

Google currently uses an index-retrieve-then-rank paradigm. Given a search query, Google comes up with a set of answers by using a mix of term-based and semantic retrieval.

These answers, or candidates, are then go through one or more re-ranking models.

The description of this process comes from a paper by Donald Metzler, Rethinking Search: Making Experts Out of Dilettantes.

How Does Google MUM Work?

Google MUM breaks away from the traditional index-retrieve-then-rank model by replacing it with a new framework.

According to a paper from December 2020, co-authored by the same Donald Metzler mentioned before, a so-called ‘Multitask Mixture of Sequential Experts’ or MoSE algorithm (source: searchenginejournal.com):

learns from the sequential order of user click and browsing data. This information allows it to model the process of complex search queries to produce satisfactory answers

In the same paper, it’s mentioned how the new algorithm could also be used to predict users’ search behaviour.

Google MUM:

  • Is based on a Transformer Architecture (like BERT)
  • Understands AND generates language
  • Is trained across 75 different languages
  • Can do many tasks at once
  • Is multimodal, which means it can understand information across text and images

To go back to the sentence above, MUM could understand that the question ‘what to do differently to prepare’ can relate to several aspects, including what gear to buy and what training programme to follow.

It could also compare the two mountains and highlight similarities and differences in their many aspects.

It could do all this by combining together information from languages different from the one used in the search query.

It could look at both text and image sources to answer the query.

MUM vs BERT

According to Google, MUM is ‘1,000 times more powerful than BERT’.

When Was Google MUM Introduced?

MUM was introduced on May 18th, 2021 in a post by Pandu Nayak, Google Fellow and Vice President of Search.

When Will MUM Be Rolled Out?

There is no official date for a roll out:

We’ll bring MUM-powered features and improvements to our products in the coming months and years

However, in an tweet exchange on May 20th, 2021 between Google’s Danny O’Sullivan and Barry Schwartz from Search Engine Roundtable, the former confirmed that they will ‘let everybody know’.

Sources

  • https://blog.google/products/search/introducing-mum/
  • https://storage.googleapis.com/pub-tools-public-publication-data/pdf/63a48017f06cd046cbf97f158195e9d96e8b2205.pdf
  • https://arxiv.org/pdf/2105.02274.pdf

Featured image by Aditya Anjagi on Unsplash

Author

Pasquale Mellone

Pasquale is Founder at Increasily.com, a H2H marketing agency based in Dublin, Ireland, and owner at print-on-demand ecommerce Mintycase.com. Pasquale has worked in Digital Marketing and Account Management since 2004. He currently lives in Dublin with his wife, stepdaughter and cat.