REPUBLIC OF YEM

UNIVERSITY OF ADEN

FACULITY OF ENGINEERING

DEPARTMENT OF INFORMATION TECHNOLOGY

GOOGLE BERT

Seminar report

Prepaed by:

Name: Haneen Hussin

Register No: 187007

Supervisor:

Dr.Shada Mabger

Feb-2023
 Acknowledgment
I thank my great god who gives me the courage to start this project in confidence
feeling. I thank Dr. Shada Mabger who helps and guide me to complete this
project.

I
 ABSTRACT

With today's technical development, online search has become an important

resource for many businesses. With the development of search algorithms in
artificial intelligence, this intelligence includes even search engines. Billions of
searches are processed every day. What's the reason? Wasn't the Google search
engine just a database where visitors searched for the information they needed?
n our time, and with the emergence of new tools and technologies every day, it has
become necessary for search engines to adjust the way they deal with page rank,
links and content in order to keep pace with these rapid changes.
Google is a real tycoon among the search engines famous all over the world. Created
in 1998 and since its incep on, the Mountain View giant has been determining
rankings based on proprietary algorithms that help provide users with the content
they are looking for.
Let's look at Google's algorithm update history over the past three years.

II
 Contents

Acknowledgment I
Abstract II
Content III
List of figures IV
Introduction 1
What are Google algorithms? 3

Why are Google algorithms constantly changing? 4

The most import Google algorithms 5

what is Google BERT? 10

How does Bert work? 11

How does the Google BERT algorithm work? 12

Why is the BERT algorithm important for improving the 13

search experience
What is the relationship between the BERT algorithm and 15
the NLP system?
Does the BERT algorithm replace RankBrain 16
Does the BERT algorithm affect SEO? 17
How can I improve SEO according to this algorithm?
Conclusion 19
References 20

III
 Table Of Figure
Fig.no Fig.name No.of.page
Fig.1 Google algorithm 3
Fig.2 Google algorithms 4
Fig.3 Panda(2011) 5
Fig.4 Hummingbird(2013) 6
Fig.5 Mobile search results 7
(2015, 2018)
Fig.6 Rank Brain (2015) 8
Fig.7 BERT (2018) 9
Fig.8 Google BERT 10
Fig.9 Bert work 11
Fig.10 improving the search 13
Fig.11 The BERT Algorithm 14
Update
Fig.12 NLP system 15
Fig.13 Search Engine 17
Optimization

IV
V
 INTERODUCATION
Nearly thirteen centuries ago, the mathematician and astronomer Abu Jaafar al-
Khwarizmi devised a method for solving mathematical problems based on following
successive steps with a specific number and each with its exact instructions.
Problems, until Google used them in our time to adjust how search results appear in
what was called: Google algorithms.
Google algorithms are considered the most important form of the rules of the game
in the search results pages.
Whoever adheres to these rules increases the odds of being ranked in the Google
search engine results, and whoever tamperes with them will not be safe from
Google punishment one way or another.
And since the word “Google algorithms” is one of the complex and mysterious
words for many, even some employees of Google itself do not understand
everything that happens in the most dominant search engine on the Internet.
If you have a website or a blog on the Internet, of course you have heard about
Google algorithms, and if you are following these developments at the moment
now, then you must have heard about the Google BERT algorithm
In general, if you have not heard before about Google algorithms or their important
role in your business, perhaps it is time now to get to know these algorithms now,
and realize their importance and role in determining the number of visits and views
that your website pages earn, which is directly related to the extent of business, and
the amount Sales maybe.
Did you feel important?
Because the search engine algorithms analyze several factors, some of which are
known and some of which are yet to be discovered, and accordingly tell Google
which pages published on the Internet deserve to appear in the first ranks of the
search results when the user writes a concept in the search engine.
The Google bert algorithm is one of the new updates that occurred in the Google
algorithms, and forms the artificial intelligence system for Google, so it is important
that you follow this article to get to know the details now and make the required
modifications to your website and benefit now.
Especially since we are witnessing that Google has launched a very big change in this
field, after its last update about five years ago, in which we previously witnessed the
emergence of the RankBrain algorithm.

1
 Initial information about the Google BERT algorithm It is estimated that this new
algorithm affects 1 out of every 10 pages on the web, and this certainly results in
dramatic changes in the order and style in which web pages will appear when you
search for a specific concept or thing on the web.

2
 What are Google algorithms?
Google algorithms are a series of algorithms that make up systems for ranking
hundreds of billions of web pages in the Google search index in order to provide
useful and relevant results in a split second. geographical.
The algorithm, as devised by Al-Khwarizmi, aims to solve mathematical problems,
but it is also used to solve problems, and the problem facing the search engine in the
Internet world is how to find the most appropriate web page for the search word,
and here comes the role of Google algorithms as they help the search engine in Find,
sort and retrieve relevant results at any time.
The algorithms work according to the criteria and rules that Google uses to show the
sites in the search results, and the matter can be likened to the fact that Google
passes the sites on several filters and accordingly chooses the sites that appear in
the advanced results, then the results that follow, and so on.

Fig. 1:Google algorithm

3
 Why are Google algorithms constantly changing?
This question has two answers.
The first simple answer is that Google algorithms change because we as humans
and societies change, and as our behavior changes, technology evolves to keep pace
with our needs and desires, and therefore search engines must also change for the
same reason. Going back about 15 years back, social media sites that help increase
website traffic did not exist, and there was no interest in website compatibility with
smartphones that were not yet widespread.
Back in our time, and with the emergence of new methods and technologies every
day, it has become necessary for search engines to adjust the way they deal with
page rank, links, and content in order to keep pace with these rapid changes.
The second composite answer is that Google, in its race with other search engines
such as: Yahoo, Bing, and others, focused on improving the user experience and
providing the best search results relevant to his request, and for that, sites that
provide useful answers to what users want were rewarded by giving them an
advanced ranking on the results pages. .
The main tool I used was to keep Google's algorithms up to date so that they hunt
down fraudulent sites with poor content, and direct attention to sites with good
content. I remember in the not too distant past, when we used Google to search for
a topic; The annoying presence of fraudulent websites in search results. For example,
when we click on the search result, we find a huge amount of ads and pop-ups that
obscure the content of the page itself, while the real content is very poor as
keywords are stacked in various forms side by side in a miserable attempt to get a
ranking. And with the advent of the Google “Panda” algorithm update in 2011 and
the updates that followed, as we shall see; The matter differed greatly, and these
disturbing results disappeared irreversibly, and were replaced by beneficial results
for the reader

Fig. 2: Google algorithms

4
 The most import Google algorithms
Passed Google algorithms with many major updates, the most important of its own
names: Penguin, Panda , Hummingbird etc.
Often focused main updates on topic or problem particular aimed at processing, so
many specialists are divided into search engines search results by period that
precedes the name and after the name.

 Panda(2011)
The main focus was on anti-plagiarism, refined content, indiscriminate mail and
abuse keywords, and pages were classified on the content quality scale, and the
grades were used as an important classification. After five years, Panda has become
part of Google's root algorithm, resulting in accelerating the implementation of
minor updates and the speed of processing the page significantly.

Fig. 3: Panda(2011)

5
  Hummingbird(2013)
This update was another major to address the abuse of keywords and low quality
content, and there was a transition from the individual keyword processing to
identify user search intentions. Instead of guessing the keywords you will get the
best results, focused on the same information and how to apply, when tackling and
arranging the ties, the importance of sympathy and similar topics and similar
searches increased significantly

Fig. 4: Hummingbird(2013)

6
  Mobile search results (2015, 2018)
This update was never given a name, although it changed the use of search in several
ways, with the primary focus being on pages without corresponding mobile versions
and site performance on mobile devices in general. There was an important shift in
ranking priorities towards pages that are well adapted for viewing and working on
mobile devices, and the optimization involved many aspects: from the size and types
of content, to how well the content is displayed on the page, and whether loading is
not blocked by external files, and so on. Even now, the performance of mobile
versions of pages remains a rather serious problem for many sites

Fig. 5: Mobile search results (2015, 2018)

7
  Rank Brain (2015)

Continuing to fight against inappropriate and low-quality content, Google has also
put a lot of emphasis on user experience with this update. Rank Brain is a grading
system built on the principle of machine learning and analysis. The update is
Hummingbird, designed to improve the quality of interpretation of user requests and
their subsequent comparison with indexed pages, and evaluation by the RankBrain
algorithm is one of the main factors behind effective ranking in shaping search
results

Fig. 6: Rank Brain (2015)

8
  BERT (2018)
The joint efforts of Panda, Hummingbird, and RankBrain are the foundation for the
next step in the fight against low-quality content, and an algorithm created by
Google applies the latest advances in natural language processing to evaluate text
content and style. The search engine is better at identifying the right keywords to
generate organic results, and the lack of context and clear topic and style become
important signals when ranking pages.
This algorithm is what we will talk about in detail

Fig. 7: BERT (2018)

9
 what is Google BERT?
In plain English, it can be used to help Google better distinguish the context of
words in search queries.
And if we want to simplify for you the role of this algorithm in the development of
Google search, we have a clear example that explains the matter.
We assume that the visitor searched for the phrase “the distance from Morocco to
Algeria.
” The phrase, as we can see, contains the words “from” and “to.” It may be The
phrase is meant to be obvious to humans but was previously less obvious to search
engines.
But with the design of the BERT algorithm, search engines are distinguishing
between these nuances to make it easier to find more relevant results

Fig. 8: Google BERT

10
 How does Bert work?
It uses the BERT Transformer, an attention mechanism that learns the contextual
relationships between words (or subwords) in a text.
In its vanilla form, Transformer includes two separate mechanisms - an encoder that
reads text input and a decoder that produces a task prediction.
Since the goal of BERT is to create a language model, only an encoding mechanism is
necessary.
The detailed workings of the switch are described in a paper by Google.
Unlike vectorial forms, which read text input sequentially (left-to-right or right-to-
left), the transformer-encoder reads the entire sequence of words at once. It is
therefore considered bi-directional, although it would be more accurate to say that it
is non-directional.
This property allows the model to learn the context of a word based on everything
around it (left and right of the word).

Fig. 9: Bert work

11
 How does the Google BERT algorithm work?
Google BERT's success has been attributed to its ability to teach a machine
natural language processing based on the full set of words in a sentence or
search operation (bidirectional training) instead of the traditional method of
training the ordered sequence of words (left-to-right or right-to-left).
The BERT algorithm allows the machine to learn the context of words based on
surrounding words rather than the word immediately preceding or following it.
That's why Google calls BERT "extremely directive binary" because contextual
representations of words start at the bottom of a deep neural network.
For example, the word "bank" will have the same representation as the word
"bank account".
Instead, the BERT algorithm creates a representation for each word based on
the other words in the sentence.
For example, in the sentence “You have entered the bank account,” the one-way
contextual form “bank” is represented based on “You have entered the account.”
While the BERT algorithm is represented for the word “bank” using both the
preceding and following context “You have entered the bank account”.
Enter.

12
 Why is the BERT algorithm important for improving the search
experience?

Fig. 10: improving the search

Now you realize that the algorithm helps Google decode human language, but
what difference does it make to a user's search experience?
It's important to remember that Google's job is to curate all content on the web
to provide the best answers for users.
For this, the search engine needs to understand what people are searching for
and what web pages are talking about. Thus, it can make the right match
between keywords and web content.
For example, when searching for “Food Bank,” the researcher understands that
the “bank” in your query does not refer to a financial institution, because
originally the word “Food Bank” meant a free service to provide food without
charge to the needy.
Even if you write misspelled “food bank” or “bank food” in reverse order, he will
also understand what you mean. With BERT, Google understands the meaning of
that word in your search terms and in the contents of indexed pages. However, in
the early days of Google, not all searches provided what the user was looking for.
The search was limited to the exact match of the keyword.
That is, when a person types “word press plugins,” for example, they are only
able to provide results for pages that use that exact term.
Since the advent of Rank Brain, Google has already begun to understand that
the word “care” is very close to the word “how to care.” Therefore, the search
engine will also display pages containing the terms “how to take care of your
health”.
The BERT algorithm makes it clear that a person wants to know how to take
care of health without sticking to exact keywords.
The problem is that Google's exact keyword-matching prototype created vices
across the internet. To appear in the search engine, many sites have begun to use
keywords in the text exactly as the user searches.
However, this makes for a very poor reading experience. Think with us, would
you rather read content that speaks naturally about the topic or a text that
repeats the keyword several times without making any sense?

13
 Therefore, it was essential that Google be able to understand search intent
which also improves the user's reading experience.
Sites are geared to producing content in natural language, using terms that make
sense to the reader. With this, Google also combats keyword stuffing, which is an
unacceptable practice that violates search engine policies.
What is meant by updating the BERT algorithm? Google’s BERT Algorithm
Update is set to change the way the search engine handles naturally formulated
searches. With billions of searches performed every day, many of them are too
subtle for a search engine to fully comprehend. This major update is designed to
allow it to process searches it can't anticipate, or to understand complex phrases
it normally can't handle.

The BERT Algorithm Update aims to assist the algorithm in natural language
processing, and the BERT AI Update aims to advance the science of language
understanding through the use of AI and machine learning in whole text.
It's something they claim represents the biggest update to their search system in
at least five years.
The biggest change in the search algorithm since the search giant's popular
RankBrain system update of 2015.
When Google announced the BERT Algorithm Update, Google VP of Search
Pandu Nayak explained how an AI-based algorithm could affect a specific search
like “Brazilian traveler to USA needs a visa.” Previously in a search like this, the
word “to” was not given any importance.
But with BERT's understanding of syntax and context, this helps the search
engine to understand that this search is specifically about Brazilian travel to the
US and not just about visas for these two countries in general. This example
shows the effect of the BERT Algorithm Update on searches, how these
advances help machine language processing and how compatible it is with
human language, and also helps to better understand what people intend to
search for when they enter search terms into the search bar or speak voice
commands to their assistants digitalists.

Fig. 11: The BERT Algorithm Update

14
 What is the relationship between the BERT algorithm and the
NLP system?
What is NLP? Neuro-Linguistic Programming, known as NLP, is a set of methods
and methods based on sensory, linguistic and perceptual principles, aiming to
develop human behavior towards excellence, creativity and development and
help people achieve better successes and achievements in their lives.
The relationship between the two terms is like a father-son relationship, where
the BERT algorithm is a descendant of the Natural Language Processing (NLP)
system.
The system belongs to a branch of artificial intelligence that deals with
linguistics, with the goal of enabling computers to understand the way humans
communicate naturally.
Examples of advances achieved with NLP include listening tools, chatbots, and
word suggestions on your smartphone.
It is well known that NLP is not a new feature for search engines.
However, the BERT algorithm represents an advancement in NLP through
bidirectional training .

Fig. 12: NLP system

15
 Does the BERT algorithm replace RankBrain?
Google is constantly studying ways to improve user experience and deliver the
best results.
This neither begins nor ends with BERT. In 2015, the search engine announced
an update that changed the search world, RankBrain.
This was the first time that Google relied on artificial intelligence algorithms to
understand content and search. Like BERT, RankBrain also uses algorithms in
computer science but does not do natural language processing.
The method focuses on analyzing searches and grouping together words and
phrases that are semantically similar but cannot understand human language on
its own. Google rambrain and bert So, when you perform a new search on
Google, RankBrain analyzes your previous searches and identifies the words
and phrases that best match that search, even if they don't match exactly or were
never searched for.
When they receive user interaction cues, the bots learn more about the
relationships between words and improve rankings.
So, this was Google's first step in understanding human language.
Even today, it is one of the ways the algorithm understands search intent and
page contents in order to provide better results to users.
Therefore, BERT Algorithm has not replaced RankBrain, but rather it is
another way to understand human language.
Depending on the research, the Google algorithm can use either of the two
methods or even a combination of the two which together form the best
computer science algorithms to provide the best response to the user.
Keep in mind that the Google algorithm is made up of a complex set of rules and
processes.
RankBrain and BERT play an important role, but they are just part of this
powerful search system.

16
 Does the BERT algorithm affect SEO?
How can I improve SEO according to this algorithm?

Fig. 13: Search Engine Optimization

What is SEO? Abbreviation Search Engine Optimization It is the science of

improving the appearance of a website or website page in the results of free
(unpaid) search engines such as Google, so that it appears in the first results, by
studying how search engines work and the factors affecting the ranking and
improvement of search engine results, as well as treating some errors and
problems. On sites, which is one of the branches of the field of e-marketing.
In view of the importance of the first search results in search engines because of
their great impact on the number of visitors to the site through search engines,
which includes the usual search and the search for images and videos (audio-
visual materials), places and news, the importance of this field has increased,
which has become specialists and experts and is even taught. .
Workers in this field are divided into two types; The owners of white hats,
which is the use of legitimate and natural methods to improve the ranking of a
website, and the owners of black hats, by using twisted, violating and illegal
methods .
As for If you are looking for the SEO steps in this article according to Google
BERT, this phrase may be disappointing.

17
 But you have to understand that Google made this update specifically to prevent
sites from optimizing pages and content for robots to focus on providing high
quality content.
The search engine wants to provide content of value to users and they want to
rely on your site for that. So don't optimize your site for Google BERT, optimize
it for your users.
That's why Google didn't give any tips for optimization, but their focus was that
we want to promote good content production practices to provide the best
visitor experience.
If you want to improve, put the visitor first, not the search engine.

18
 Conclusion
According to what I read in this important article, the preparation of useful, clear
and deep content forms the basis for your site or blog to occupy the first ranks in
Google.
In conclusion, what Google certainly seeks is to serve the browser and the user,
and help them find results that meet their needs and solve their questions and
problems. .
Google also adds a set of policies every once in a while, and here are the most
important of these policies that Google follows to fight Internet pirates to
provide its best services to users with every new update from Google
1. Anonymity: It is a policy that prevents deception, so that the title of the page in
Google is different from its content on the site.
2. Portals: It is another type of deception, for example, if you own a group of sites
with a slight difference in their domain address in order to rank higher in
Google.
3. Hacked content: Some WordPress sites are subject to hacks. These processes
create a very common problem. Google added Japanese pages, which are policies
that prevent this content from appearing.
4. Hidden text and hidden links: Imagine that you are writing in white on a white
background, this is exactly the hidden text. Any practices that hide text from the
user on your website, and try to show it to search spiders only are known as
Black Hat SEO strategies.
5.Crowding of keywords: In the past, keywords should have been repeated strictly
only for ranking, after Google updated its algorithms, this is now impossible.
6. Spam links: Any link that has no value is considered spam, the link must
provide value to the user, when they open it on their browser, Chrome browser
should face to a relevant page.
7. Automated Traffic: Some bots work on Google Chrome specifically to increase
website traffic with the aim of improving rankings, no longer .
8 .Malware and Harmful Behaviors: Since Google has been fighting viruses,
Google wants to keep its programs safe.
9. Misleading functions: The new version of the new Google update blocks any
shady sites, for example, sites that claim to provide Google Play Store credit but
do not, or sites that claim to offer Google Chrome browser update but this is not
true.
10. Stolen content: Google warned of theft in its latest update. Google does not
allow any theft operations at all on its engine, so Google prevents sites that copy.
Core updates are changes made by Google to improve search in general and
keep pace with the changing nature of the web. Knowing that core updates are
not targeted in any way to specific websites, these updates may result in some
noticeable changes in the performance of websites

19
 Reference
https://seo-master.marketing/search-engine-
algorithms/?fbclid=IwAR3fU_qj5geQgFd HKkFIerRwxJwxtN0EkvxKY6W2dM1MbpXqMDAsx_
Q0E

h ps://www.google.com/url?sa=t&source=web&rct=j&url=h ps%3A%2F%2Fwww.seo-
ar.net%2F%25D8%25AC%25D9%2588%25D8%25AC%25D9%2584-
%25D8%25AA%25D8%25B7%25D9%2584%25D9%2582-
%25D8%25AA%25D8%25AD%25D8%25AF%25D9%258A%25D8%25AB-
%25D8%25A8%25D8%25B1%25D8%25AA-bert-
%25D8%25A5%25D9%2584%25D9%258A%25D9%2583-%25D9%2583%25D9%2584-
%25D9%2585%25D8%25A7-
%25D9%2586%25D8%25B9%25D8%25B1%25D9%2581%25D9%2587-
%25D8%25B9%25D9%2586%2F%3Famp&ved=2ahUKEwjQxqjT08X9AhX5TKQEHa2SB2cQFno
ECCAQAQ&usg=AOvVaw3y7Ej_yn2tZ0Iye96FJ-
zT& clid=IwAR0mrP4pa3NSuGbMBFcvevxj7fIZlN0hdHhsy5QFq1B8H7JxA5nHy1vJD7s

h ps://www.google.com/url?sa=t&source=web&rct=j&url=h ps%3A%2F%2Fwww.simplilea
rn.com%2Fbert-the-new-google-algorithm-
ar cle&ved=2ahUKEwjdw6q81MX9AhViaqQEHfd6DoY4HhAWegQIBBAB&usg=AOvVaw0dE7l
v4gbDoLSyqDZWqULQ& clid=IwAR1GDrzW3XHBzD5KT1CnKDRFLLaXoWmDJajdyCJQzoEed9
ziDQY2rDy1fM8

https://www.google.com/url?sa=t&source=web&rct=j&url=h ps%3A%2F%2Flearn.microsof
t.com%2Far-sa%2Fazure%2Fmachine-learning%2Fhow-to-configure-auto-
features&ved=2ahUKEwjupaah1MX9AhXqTKQEHcKvAYk4HhAWegQIBBAB&usg=AOvVaw0qP
a7bX0LT-
Ev1s4QDfZ7i& clid=IwAR1_V6GAqsR7Kq0lM0KpIPk4mOeG1f8eH9MVPwb7rSvWSbHngmAh
XGqkH-U

h ps://www.google.com/url?sa=t&source=web&rct=j&url=h ps%3A%2F%2Fwww.simplilea
rn.com%2Fbert-the-new-google-algorithm-
ar cle&ved=2ahUKEwjdw6q81MX9AhViaqQEHfd6DoY4HhAWegQIBBAB&usg=AOvVaw0dE7l
v4gbDoLSyqDZWqULQ& clid=IwAR0oKt0NLZU8GVXXyQUHB8X3cm61DwCG-
IfKC_aynp2Y7aslhQsJucw-2qI

h ps://www.google.com/url?sa=t&source=web&rct=j&url=h ps%3A%2F%2Fwww.geeksfor
geeks.org%2Fexplana on-of-bert-model-
nlp%2F&ved=2ahUKEwiTjsOH1MX9AhVcaqQEHaBmCY04ChAWegQICxAB&usg=AOvVaw0f6l
nfC8marM2_PO3kXIPU& clid=IwAR3vNuUmLM2VEkd4_2FjKbJ7soAm1xmyEPW87GvCfgFRb
khNODG798jZosc

20