About Google and its semantic understanding
Posted: Tue Feb 18, 2025 6:45 am
Google's goal has always been to return the best possible results for a user's search . Over time, it has improved aspects of semantics, which allow it to better understand both the meaning of the user's search and the content of websites. Through the improvements that Google has been applying with its algorithms , SEO practices have also been changing. In the past, to position a keyword of medium difficulty, it was enough to do some on-page SEO , a lot of keyword stuffing , some links (or not even that) and that was it. Now that is not only insufficient, but it is clearly contrary to several guidelines for positioning in its search engine.
Nowadays, when we perform a search, Google takes into account hundreds of aspects : my possible search intention, my browsing history, my geolocation, the language in which I searched, related entities, current aspects and many more parameters, and returns a results page that may include blue links but also multiple content formats.
Summary of contents:
NLP and its deployment in algorithms and results
NLP, Entities and Knowledge Graph
The Google NLP API
How Google interprets semantics
HTML and DOM
Anchoring and semantics
Other factors that Google takes into account
How we target our content to certain searches
Semantic analysis of a content
NLP and its deployment in algorithms and results
NLP stands for Natural Language Processing. Applied to SEO , you should understand that Google uses it to better understand the user's search , and the content of the sites, in order to make the desired match. When we put a query in the search engine that contains a set of words, Google performs a semantic analysis, detects relationships between the words, and obtains a joint malta whatsapp data meaning of the search , as well as the related entities. Google also uses it to understand and relate data about current events, news, or even conversations in forums or social networks about the subject.
Tokenization is a key process to understand how NLP works at Google: this concept, normally applied in blockchain jargon , in this case refers to how it separates words in a text , analyzes them separately and as a whole, and in turn with the possible punctuation marks applied.
Another key to the process is grammatical categorization . Do you remember when we analyzed sentences at school, separating phrases and direct, indirect objects, etc.? Well, it's a very similar process. There you label nouns, adverbs, verbs, prepositions, and understand what nuances or changes each part of the sentence implies in the meaning of the whole.
Another dimension to consider is lemmatization . Through this process we transform a word into the concept that we would find in the dictionary. If in a sentence we find the concept finalized , for example, we lemmatize it with the dictionary entry for the verb finalize . Let's see a very clear example: we search for “SEO positioning”, and through lemmatization, it returns a result from Wikipedia referring to “search engine positioning”, which would be a more technical concept:SEO lemmatizations
The next concept you should familiarize yourself with is dependencies. There is no doubt that some terms depend on others according to grammatical rules. If I say “ this post is very good ”, Google should understand that good belongs to post, not to the word immediately before it, as it processed previously.
NLP, Entities and Knowledge Graph
All this is good to note, as it is related to a phenomenon that we already know: the Knowledge Graph . These special results that we see in the SERPs actually draw on a key concept in SEO today: entities. These are objects or concepts that do not give rise to error , and that establish relationships with keywords. In searches related to “Barcelona”, our city, a clear entity would be the city as a geographical element, but Google (and users) will need to differentiate it from another entity that is just as or better known, the football club. Even the legendary president of Barça Josep Lluís Núñez referred to the city of Barcelona as “the city that bears the name of our club”. Let’s see what happens if we search for “Barcelona”:
NPL SEO
But if we add a nuance, it already identifies that it is another entity, and by its nature the type of results and information will be different:
SEO knowledge GRaph
If you want to know which entities a content is related to, for possible analysis or further improvements, you can use Inlinks' entity indexing checker . It is a simulator of what we know as salience , a scale that indicates the importance (from 0 to 1) of an entity in the analyzed document. We have provided the tool with our SEO dictionary , and the result is the following:
entity indexing checker
You will see that it tells us that Google probably detects two entities, in this case SEO and dictionary, and this is correct. It also gives us guidelines to better identify other related entities.
The Google NLP API
We present to you a very powerful tool, which is also free to a certain extent: Cloud Natural Language . You can pass it texts, and it tells you how Google analyzes that content, and what entities it extracts . To give you a slightly different example, I have passed it my post on Open Graph and the result is the following:
Nowadays, when we perform a search, Google takes into account hundreds of aspects : my possible search intention, my browsing history, my geolocation, the language in which I searched, related entities, current aspects and many more parameters, and returns a results page that may include blue links but also multiple content formats.
Summary of contents:
NLP and its deployment in algorithms and results
NLP, Entities and Knowledge Graph
The Google NLP API
How Google interprets semantics
HTML and DOM
Anchoring and semantics
Other factors that Google takes into account
How we target our content to certain searches
Semantic analysis of a content
NLP and its deployment in algorithms and results
NLP stands for Natural Language Processing. Applied to SEO , you should understand that Google uses it to better understand the user's search , and the content of the sites, in order to make the desired match. When we put a query in the search engine that contains a set of words, Google performs a semantic analysis, detects relationships between the words, and obtains a joint malta whatsapp data meaning of the search , as well as the related entities. Google also uses it to understand and relate data about current events, news, or even conversations in forums or social networks about the subject.
Tokenization is a key process to understand how NLP works at Google: this concept, normally applied in blockchain jargon , in this case refers to how it separates words in a text , analyzes them separately and as a whole, and in turn with the possible punctuation marks applied.
Another key to the process is grammatical categorization . Do you remember when we analyzed sentences at school, separating phrases and direct, indirect objects, etc.? Well, it's a very similar process. There you label nouns, adverbs, verbs, prepositions, and understand what nuances or changes each part of the sentence implies in the meaning of the whole.
Another dimension to consider is lemmatization . Through this process we transform a word into the concept that we would find in the dictionary. If in a sentence we find the concept finalized , for example, we lemmatize it with the dictionary entry for the verb finalize . Let's see a very clear example: we search for “SEO positioning”, and through lemmatization, it returns a result from Wikipedia referring to “search engine positioning”, which would be a more technical concept:SEO lemmatizations
The next concept you should familiarize yourself with is dependencies. There is no doubt that some terms depend on others according to grammatical rules. If I say “ this post is very good ”, Google should understand that good belongs to post, not to the word immediately before it, as it processed previously.
NLP, Entities and Knowledge Graph
All this is good to note, as it is related to a phenomenon that we already know: the Knowledge Graph . These special results that we see in the SERPs actually draw on a key concept in SEO today: entities. These are objects or concepts that do not give rise to error , and that establish relationships with keywords. In searches related to “Barcelona”, our city, a clear entity would be the city as a geographical element, but Google (and users) will need to differentiate it from another entity that is just as or better known, the football club. Even the legendary president of Barça Josep Lluís Núñez referred to the city of Barcelona as “the city that bears the name of our club”. Let’s see what happens if we search for “Barcelona”:
NPL SEO
But if we add a nuance, it already identifies that it is another entity, and by its nature the type of results and information will be different:
SEO knowledge GRaph
If you want to know which entities a content is related to, for possible analysis or further improvements, you can use Inlinks' entity indexing checker . It is a simulator of what we know as salience , a scale that indicates the importance (from 0 to 1) of an entity in the analyzed document. We have provided the tool with our SEO dictionary , and the result is the following:
entity indexing checker
You will see that it tells us that Google probably detects two entities, in this case SEO and dictionary, and this is correct. It also gives us guidelines to better identify other related entities.
The Google NLP API
We present to you a very powerful tool, which is also free to a certain extent: Cloud Natural Language . You can pass it texts, and it tells you how Google analyzes that content, and what entities it extracts . To give you a slightly different example, I have passed it my post on Open Graph and the result is the following: