1. Home
  2. /
  3. EDGE of the Web
  4. /
  5. Understanding Search Ontology and Google’s Language Models of the...

Understanding Search Ontology and Google’s Language Models of the Past and Future | Bill Slawski

By Site Strategics
July 6, 2021

Search ontology and taxonomies can help you better understand how Google looks at content and the relationship between words and phrases. Language models are used to examine the content. When you’re able to understand how language models examine the content and understand user intent and context, you’re able to provide the user with exactly what they’re looking for. Search engines want to answer a user’s questions, and you want to give the search engine that answer.

Search Ontology and Taxonomies

Taxonomies and Ontology are the two main players when it comes to searching. Taxonomy organizes content into certain categories, while ontology tells you how content is related and focuses on relationships between content. Ontology is more sophisticated than taxonomy. 

With taxonomy, broader and more specific content is categorized on levels. The content is broken down, categorizing it by what it does for the content. Taxonomy takes on the structure of the content, its organization, and its navigation. 

Ontology focuses on the relationship between content. It is attempting to understand the content while mapping the relationship one word or phrase has with another. Google’s knowledge panel is continually trying to understand the mapping of relationships because, to Google, relationships and knowledge are synonymous with each other. Google can rely on a content relationship model that maps the relationship between content and the value of that content according to the outside world.    

What are Language Models?

Language models are trained to examine the content in a certain way. This means that language models can discern entities on a page by searching for places, names, proper nouns, and fact patterns related to these. Language models are actively examining human behavior and learning and adapting to it.  

Language models can build Google’s knowledge graphs which explain why certain things have a relationship. What a language model creates, known as “triples,” is built with a subject, verb, and object. Looking at knowledge graphs, content strategists can then fill any gaps to make the relationships even stronger.      

Looking at Past Language Models

BERT, also known as bidirectional encoder representations from transformers, is Google’s language model that would take the number of words and phrases on a page and learn about language and content based on these. BERT was able to take words and syllables on the page and, through statistics, establish the type of word and syllable present on the page. This mainly determines any given site’s quality in comparison to other sites. 

BERT could take words being used to determine a relationship within content as well as the user intent. BERT’s abilities were based on the human mind’s ability to recognize entities, such as people, places, and things, within objects in our world. 

Google’s unified model of understanding, the Multitask Unified Model, or MUM, can connect words with pictures or products. MUM also examines formats which allows it to see how something might be best displayed on a page. 

Through MUM, Google matches many different formats, not just the standard blue link to a user’s search. Due to MUM being relatively newer, there is still room for evolution—of both the model and the user’s intent. MUM also follows through to determine whether the user used the information given or not.   

The Connection Between Context and User Intent

Context is essential for anyone searching on Google, and it is just as important to those creating content as language models learn how to read the context. With the importance of context comes the ability of a language model to understand user intent. 

BERT, being able to process language and content based on the words and syllables on a page, allowed Google to recognize action items. The language model can then determine the number of times certain words show up on the page and interpret their relevancy. 

Language models take the content and find relationships between multiple pages. The context of the page shows Google what that page is about. It’s important to understand the context and how Google interprets context and incorporate that instead of solely focusing on keywords. 

Google’s language models additionally attempt to see how certain words work together to form phrases. The language models no longer look for foundational associations and instead focus on the contextual elements of the page. This allows Google to provide the user what it is looking for and better understand how to determine what the user wants to use certain information or. 

Context allows Google to determine what it is that the user is looking for. There are a lot of words that, given the context, can be interpreted in different ways. Language models understanding context enables them to discover user intent. 

The Search Engine’s Purpose

The search engine’s job is to fill the user’s needs, whether this is informational or situational. A Google bot reads and organizes pages to make something more findable for the user. Google has realized that a user doesn’t just want a list of blue links; they want their question answered right then and there. 

A search engine is a question answered and gives the user what they want and predicts what it thinks the user will do next. The search engine’s sole purpose is to fulfill the user’s needs and offer as much to the user as possible for them to get the most out of their interaction.  

The Takeaway for SEO Strategists 

Due to Google making connections for the user, Google will favor you more if you focus on relationships between concept and entity rather than just SEO. SEO strategists should be trying to fill in the gaps for the user. You want to be as helpful as possible and provide solutions for the user. 

The goal of a page is to answer the user’s question. SEO strategists should strive to have the answer and make it appear on the page while providing all the information the user may not even know they need or want. Google’s language models, along with ontology and taxonomies, will provide the user exactly what they want, so you want to fit your page to what Google is looking for. 

Connect with Bill Slawski

Twitter: @bill_slawski (https://twitter.com/bill_slawski)

Facebook: https://www.facebook.com/bill.slawski

LinkedIn: https://www.linkedin.com/in/slawski/

Connect with Erin Sparks, Host of EDGE of the Web and CEO of Site Strategics

Twitter: @ErinSparks (https://twitter.com/erinsparks)