Google Search is able to understand human language with the help of multiple AI models that work together to find the most relevant results.
In a new post on the company’s official blog, Pandu Nayak, Google’s vice president of search, briefly explained how these AI models work.
Nayak demystifies the following AI models that play an important role in how Google returns search results:
- RankBrain
- neural matching
- BERT
- mother
None of these models work alone. They all help each other by performing different tasks to understand queries and match them to what searchers are looking for.
Here’s a look at Google’s key takeaways behind the scenes, a look at what its AI models do and how it translates into better results for searchers.
Google’s AI model explained
RankBrain
Google’s first AI system, RankBrain, was launched in 2015.
As the name suggests, the purpose of RankBrain is to find the best order for search results by ranking them according to relevance.
Despite being Google’s first deep learning model, RankBrain continues to play a major role in search results today.
RankBrain helps Google understand how words in search queries relate to real-world concepts.
Nayak explains how RankBrain works:
“For example, if you search for ‘what is the title of the consumer at the top of the food chain’, our system learns by seeing these words on different pages that the concept of the food chain may be related to animals and not human consumers.
By understanding these words and matching them to their related concepts, RankBrain understands that you’re looking for what is commonly referred to as a “top predator.”
neural matching
Google introduced neural matching to search results in 2018.
Neural matching allows Google to use broader conceptual knowledge to understand how queries relate to pages.
Rather than looking at individual keywords, neural matching examines entire queries and pages to identify the concepts they represent.
With this AI model, Google is able to cast a wider net as we scan its index for content relevant to a query.
Nayak explains how neural matching works:
“Take the search for ‘insight on how to manage greens’. If a friend asked you this, you might be stumped. But with neural matching, we were able to understand it.
By looking at the broader representation of concepts in the query—management, leadership, personality, etc.—neural matching can decipher what management skills the searcher is looking for based on popular, color-based personality guidelines. “
Screenshot from blog.google/products/search/, February 2022BERT
BERT was first introduced in 2019 and is now used for all queries.
It is designed to do two things – retrieve relevant content and rank it.
BERT understands how words relate to each other when used in a specific order, ensuring that important words are not excluded from the query.
This sophisticated understanding of language enables BERT to rank the relevance of web content faster than other AI models.
Nayak explains how BERT works in practice:
“For example, if you search for ‘can you get drug for someone pharmacy’, BERT will understand you’re trying to figure out if you can get someone else’s pharmacy.
Before BERT, we took this short preposition for granted, mostly sharing results on how to prescribe. Thanks to BERT, we understand that even small words can have great meaning. “
Screenshot from blog.google/products/search/, February 2022mother
Google’s latest AI milestone in search, the Multitasking Unified Model (MUM), launches in 2021.
MUM is a thousand times more powerful than BERT and can understand and generate language.
It has a more comprehensive understanding of information and world knowledge while being trained in 75 languages and many different tasks.
MUM’s understanding of language will span the future of images, text, and more. When you hear MUM called “multimodal,” that’s what it means.
Google is in the early stages of realizing the potential of MUM, so its use in search is limited.
Currently, MUM is used to improve searches for COVID-19 vaccine information. In the coming months, it will be available in Google Lens as a way to search using a combination of text and images.
generalize
Here’s a review of Google’s main AI systems and what they do:
- RankBrain: Rank content by understanding how keywords relate to real-world concepts.
- neural matching: Provides Google with a broader understanding of concepts, thereby expanding the amount of content Google can search.
- BERT: Allows Google to understand how words change the meaning of a query when used in a specific order.
- mother: Understand information and world knowledge in dozens of languages and in many forms, such as text and images.
These AI systems work together to find and rank the most relevant content for a query as quickly as possible.
source: Google
Featured image: Igor Golovniov/Shutterstock
