What is Word2Vec? A Simple Explanation | Deep Learning Tutorial 41 (Tensorflow, Keras & Python)

[@codebasics]

Check out our premium machine learning course with 2 Industry projects: codebasics.io/courses/machine-learning-for-data-science-beginners-to-advanced

[@mehmetbakideniz]

I started searching wordtovec videos after failing to understand it by following NG's lessons. That is the single video that can actually tell that the word embeddings are 'the side effects' of the training process and this is how it finally clicked for me. Thank you very much!

[@anubhavthakur2985]

then you didn't searchthe youtube enough

[@ahmedgamberli2250]

I like how you love your homeland and use it in all examples. Greetings and Love from Azerbaijan.

[@sunilgundrai6464]

As part of my NLP dissertation, I was looking for some real time use cases with some clear explanation. I found this a super useful and thank you for great demonstration with so many examples which are easy to understand. You rock with your teaching skills!!

[@girmayohannis4659]

nice to meet you here, in which university are you studying for your PhD?Thanks

[@austinburcham4260]

This was unbelievably helpful. This filled in the intuition gap behind word2vec that I was so desperately looking for.

[@assafbotzer7952]

So clear, so eloquent, and so concise. Your contents are gift to this world. Thank you for using your intelligence, diligence and teaching skills to make a positive mark.

[@fidaeharchli4590]

i confirm

[@raom2127]

Presenting complex understand matter in an simplified way Dhaval Sir we you are an patience,consistent, simplified ,organised way of subject presentation expert.Basics-Theroy-Coding-Pratice..with..Great Explaination.

[@sanjeebkumargouda1471]

Great explanation .. 🙌🙌🙌 After watching many videos on this topic finally my understanding is cristal clear. You are doing awesome job sir.

[@codebasics]

I appreciate you leaving a comment of appreciation

[@mimansamaheshwari4664]

One of the best videos on word2vec

[@gayathrigirishnair7405]

This is the video that finally helped me grasp this concept. Thank You!

[@robertcormia7970]

This was a useful introduction, I don't have the math chops to understand it, but it was useful to hear some of these definitions.

[@sumit121285]

you are the real teacher.....what should i say for you ???? thank you sir...thank you so much.........

[@anonymous-or9pw]

He played it really well when he marked male = -1

[@yonahcitron226]

incredible content. this guy is one of the best on youtube

[@codebasics]

I appreciate you leaving a comment of appreciation

[@Sunilgayakawad]

Crystal clear explanation!! Thanks you so much sir

[@vgreddysaragada]

Super explanation ..Thank you so much

[@SoftRelaxingAndCalmMusicNature]

Well done. This is one of the best course on word2vec so far. I do have a master degree in AI and event that I did not work professionaly in the field your cour brough a lot of souvenirs haaa.. During my master 15 years ago I introduced an archaich method for resolving question/answering based on linkgramar, wordnet, verbnet and semnet. At the end of my syntactical analysis I also discovered that by just using world context it was possible to comme up with a vector representation of named entities.. The innovation is here is the use of neural network to give a value to the world. This is just brilliant. In my thesis I was already showing that language is just a code representing a subjective version of one universe and that human and animal comunicate using theirs own code.

[@PremKumar136]

Awesome explanation. Crystal Clear.

[@kanisrini01]

Amazing Video 👏🌟. Thank you so much for the great explanation

[@619vijay]

Very useful and informative

[@farnoushnazary7295]

Awesome explanation. Thanks!

[@overthrowenjoyer]

What a great video and what a great explanation! ❤

[@moni1122331]

great teacher, great explanation, great presentation, great context

[@fidaeharchli4590]

thank you

[@tagoreji2143]

Good Explanation Sir.Thank you

[@kimdaeeun6683]

Easy explanation!! Tks much👍👍

[@ledinhanhtan]

Mind blowing 🤯🤯 Thank you!

[@javierlopezcampoy5951]

Great explanation! Thank you very much

[@notknown9307]

thanx we are learning a lot from you

[@codebasics]

Glad it was helpful!

[@notknown9307]

@@codebasics waiting for your next upload you are doing your work very well👍👍

[@dhirajkumarsahu999]

Great Visual way of teaching! Thank you so much Sir ❤️

[@shivav7379]

A Very good explaination - really very helpful

[@moutazalreesh1763]

Thank you so much for this amazing explanation! . I was wondering if it's possible to download the PowerPoint slides used in this course. Could you please let me know where I can access them?

[@abir95571]

There's a subtle mistake in your CBOW explanation at 8:34 . In CBOW the target is always the central word based on context i.e the surrounding word . That means for a substring "Emperor ordered his" and window size of 3 the target is "ordered" and features are "Emperor , this"

[@ashwinshetgaonkar6329]

so he explained skip gram

[@ashwinshetgaonkar6329]

so he explained skip gram

[@abir95571]

@@ashwinshetgaonkar6329 yes

[@BARaaz04]

Very good explanation. Thanks.

[@namansethi1767]

Thank you Sir for this playlist

[@neerajashish7042]

Approximately how many videos are going to come in this series except the existing videos, by the way thanks a lot sir, the only playlist on youtube which was way more knowledgeable for machine learning and deep learning..

[@codebasics]

There will be atleast 5 to 10 videos coming up and then I will start the project series

[@ashwinivalmiki7636]

Hello sir, Please make a video on GRE and IELTS preparation , this will be more useful and helpful to students like me planning to study Masters Abroad as your videos are clear, we get motivated . Thank you.

[@vishaldas6346]

Also what would be your next topic in deep learning, is it sequence to sequence models?

[@ShahabShokouhi]

I was watching Andrew Ng's course on sequence models and his lecture on word2vec is just a bullshit. Thanks god I found your video, amazing explanation.

[@ashokkonatham8857]

Wow, very very clear . Thank you 🙏

[@codebasics]

Glad it was helpful!

[@vinaykumardaivajna5260]

Great explanation as always

[@list10001]

Thank you! The explanation was very clear.

[@shubhamwaingade4144]

Awesome explanation of the concept!

[@pradeept328]

Great explanation

[@vishaldas6346]

I think Dhaval, there is no non-linear activation function between the input layer and hidden layer. Correct me if I am wrong.

[@darshangangurde7855]

thanks a lot...holly great..pls complete the playlist asap

[@manikant1990]

superbly explained !!

[@vikaspatildod]

Beautiful video

[@minruili4789]

Fantastic explanation!

[@pretomghosh6231]

This COBW and Skipgram is kind of encoding+decoding architecture like the autoencoders if I am not wrong?

[@notknown9307]

Excited 😄

[@madhu1987ful]

Awesome man...loved it...can you pls upload some code walk through of this concept -- some gud projects

[@harshvardhanagrawal]

Where do we get the predicted output from? How do we enter it for comparison?

[@phil97n]

Awesome thank you

[@taufiqulhaque4987]

would you please create a playlist on NLP?

[@lohitsalavadhi6912]

Great explained finally

[@codebasics]

🙏🙏

[@trendyjewellery1987]

Superb

[@wenzhang5879]

I think you mean 'side products' rather than 'side effect'?

[@BeradinhoYilmaz]

is there standart real list for every onject given here. For example for cats, tails 0.2?

[@wp1300]

7:20 Meaning of word can be inferred by surrounding words

[@bibhupadhy4155]

Great Explanation :) Crisp and to the point , Better than Hrithik Roshan Super Hero Movie's Explanation :P :P

[@ChaitraC9191]

Hello I have doubt in this explanation, aren't all the weights gonna be same when our neural network is trained ? what I mean is once we train a network W(T)X is what triggers a output node so how do we have different weights for every output word

[@cherupawan3777]

Did u got answer to this

[@djelloulbouchiha-cunaamaal7848]

We need a course about NLP Transformers..

[@thurakyawnyein6113]

superb..

[@Cooldude5786]

The statement "King - man + woman = Queen" is well-known in machine learning. However, when we examine the characteristics of a king, they often include being super rich, having authority, and possibly not having a tail. Yet, there is a contradiction: a lion is also referred to as a king, and it does have a tail. How can a computer differentiate between a human king and an animal king? Doesn't this introduce bias since the training corpus typically associates "king" with humans rather than animals? Just because something appears less frequently or is absent from the corpus doesn't mean it lacks value or significance.

[@lisali6205]

you are the best

[@imanqoly]

The more you dig deeper into a thing, the greater the tutor gets

[@amanagrawal4198]

I think there's a mistake , because in both cbow and skip gram , the weights that made the embdeddings are always between input and hidden layer , and here in cbow you mentioned the weights between hidden and output are considered.

[@amanagrawal4198]

CBOW Model Architecture Review The CBOW architecture works by predicting a target word based on the context words around it. Here’s a step-by-step explanation of the flow: Input Layer: This consists of several one-hot encoded vectors corresponding to the context words. Projection Layer (Hidden Layer): Each one-hot vector is used to retrieve a word embedding from the first weight matrix (input-to-hidden weights). Unlike typical neural networks, there is no activation function here; the embeddings are simply summed or averaged (depending on the implementation) to produce a single dense vector. This vector represents the combined semantic content of the context words. Output Layer: The averaged embedding vector is then projected to the output layer using a second set of weights (hidden-to-output weights). The output layer is a softmax layer that predicts the probability distribution over the vocabulary for the target word. Role of Weights in Embedding Formation Input-to-Hidden Weights: This is essentially the embedding matrix. Each row in this matrix corresponds to the embedding of a word in the vocabulary. When context words are fed into the model, their embeddings are retrieved by indexing this matrix with the one-hot vectors. These embeddings are what you typically extract and use as pre-trained embeddings for other tasks. Hidden-to-Output Weights: These weights are used to transform the combined embedding from the hidden layer into a prediction for the target word. Each column in this matrix (since it's typically the transpose of the embedding matrix in many implementations) can be seen as a "contextual embedding" of a word when it acts as a target.

[@yasaswinigollapally7603]

Sir your video is awesome 🙌,i have one doubt ,what is the main difference between skip gram and bag of words model?

[@prasanth123cet]

Will be get nearly identical word vectors for CBOW and skim gram methods for a particular word say 'king'?

[@thamizharasim5970]

Thanks a lot 😌

[@rahulsoni412]

Thanks a lot for explaining this using a neural network diagram :)

[@codebasics]

🙂👍

[@rahulsoni412]

@@codebasics can you explain how the number of weights are calculated in word embedding, I mean the number of total weights. I was getting confused while calculating the number of weights.

[@prasannan-robots]

Thanks for this awesome tutorial waiting for coding part :)

[@anpowersoftpowersoft]

Amazing

[@uwaisahamedimad556]

Hi, it is a wonderful explanation for word2vec I've ever seen.I have a question,I have my own corpus and I have built multiple wor2vec models, How to evaluate these models and how am I gonna choose the best one???

[@codebasics]

One approach is to take a classification or some other NLP problem in your domain and build NLP classification model using your embeddings. You can then check the performance of those models to evaluate how effective embeddings are

[@uwaisahamedimad556]

@@codebasics thanks a lot for the reply. based on your answer it seems like there is no standard or at least a well-established evaluation method for the performance of word embeddings.

[@jongcheulkim7284]

Thank you.

[@lemoniall6553]

Is word2vec using dimensional reduction too?

[@anandakhan4160]

sir, how do u unzip the json file using git bash , is not clear to me. help me plz. thanks.

[@arjunbali2079]

Thanks sir

[@kmnm9463]

Hi Dhaval, Great video on W2V, The link for the coding part of implementing Word2Vec in Python, please?

[@codebasics]

Yes that video is coming up soon. I have not yet uploaded it

[@bii710]

That was a great explanation. Thanks. I have this one question in my mind. If all words in documents are unique then how word2vec will find vector for the last 2 words? Considering cbow

[@vitocorleone1991]

Brilliant

[@akshansh_00]

bam! life saver

[@umerfarooque6373]

How to evaluate a word2vector model

[@ibrahemnasser2744]

What a mathematician would do when he/she hear you say "a vector is nothing but a set of numbers"

[@sebinsaji9573]

Can you say about cyber security scopes skills

[@amanbajaj7591]

wheere is neural network link?

[@houchj0372]

doesn't CBOW mean Contextual Bag of Words?

[@codebasics]

Continuous Bag Of Words: analyticsindiamag.com/the-continuous-bag-of-words-cbow-model-in-nlp-hands-on-implementation-with-codes/

[@houchj0372]

@@codebasics you are correct, thank you. By the way, this video is excellent.

[@PavanKumar-bk1sz]

Can I get an admission in bsc data science after 12th commerce in St Xavier's College Mumbai ???? and I've mathmatics in optional subject ??? please please please please please please please please please tell me I've been requesting you for 6 months 🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏🙏

[@humanardaki7911]

working?

[@johnnysaikia2439]

King of the jungle has tail though

[@amitmishra5474]

Lion King has tail 😅

[@mmenjic]

3:56 why horse and woman are same gender for start ????? then king minus men is gender -2 adding a woman or horse to that you get gender -1 which is men or king !?????

[@priyeshsrivastava8025]

no its (-1) - (-1) + (+1) = +1 i.e. queen

[@pawansinha8442]

but in case of king of jungle that is lion, he has a tail,😃 just saying...

[@RePuLseHQKing]

3:35 paygap lmao

[@santoshsaklani5019]

Kindly make video on vulnerability prediction using wordtovec

[@mubashiraqeel9332]

the thing is your all videos are connected to previous I am unable to watch a whole video you always made me pause and watch a previous video that's really a problem first i was watching the text classification video you said go watch bert first then in that video you said go watch word2vec then you said go watch part 1 first then now in this video you said go watch neural network now tell do you really want me to watch a whole video because i am just opening a new tab repitively.

[@danielbrockerttravel]

This headline is a complete lie that is not how embedding work and it’s highly unethical

[@greenweed3253]

can you recommend other sources then?

[@danielbrockerttravel]

@@greenweed3253 Statquest teaches it more responsibly. He specifically told me he avoided the king-man + woman = Queen example because it simply isn't how word vectors function in practice.

[@shivangiawasthi9388]

found a better explanation here - kzbin.info/www/bejne/gJ7Ik5SXpaaWgc0

[@PavanTripathi-rj7bd]

Great explanation!

[@debatradas9268]

thank you so much