I like your videos first and then start watching your Data Science videos because I am sure that after I am done watching it, I will like it anyway. Keep it up.. 🙏
if the word 'healthcare' did occur in all 3 speeches, but occurs in the Obama speech 26 times, but only once in Clinton's and Bush's speeches. Using this mechanism, the IDF of healthcare would still be 0, but since the word has been used a considerably large number of times in the Obama speech, it is definitely important
For any given word/term, we want to know how important is that term for a given document, relative to the entire corpus of documents. E.g. for Clinton these subset of words is really important in his inauguration speech, relative to the other inaguration speeches. TF-IDF is simply a multiplication of the metrics TF (term frequency) and IDF (inverse document frequency).
You saved me! My professor explained this in 3 hours, I watched it 2 times and I don't get it. This guy explained the same concept in 7 minutes and I get it!
How do you model multiple objects associated to a term class: Dental Care: United Health Care, Blue Shield, ..., by state? This becomes contextual and local within the text - how close is the word dental care in the text to UHC, for instance. The result would show which states address dental care in their health insurance regulations and which insurance companies make it available - both in a positive and negative way. Understand that this is a narrow example. Thanks
Hi, first of all, thanks for the great explanation. I have watched your videos about Word2Vec and TF-IDF, and I need help, please. I'm a student working on a project about binary classification of SQL injection attacks. The dataset I have contains two columns: 'sentence' and 'label.' I need to extract features, but I'm confused about which technique to use: Word2Vec or TF-IDF. Can you help me decide?
Is this a good tool to create a top of "important" words in a dataset? or it just helps to see the relevance in a particular document, I want to use it so I can maybe sum all the tdidf of all the documents and create a top words but I don't know if this is the best approach/solution to what I want, thank you in advance
This is a great explanation. Thanks. I have a question about differences between the implementation described in this video and another implementation commonly found on the web. Can you explain how these two details would impact the final representation: 1) Term frequency simply calculated as term count 2) Applying vector normalisation (L2) to the document vector obtained in this video Another question which is more open-ended: why is TfIdf still relevant ? Or less provocatively - is there a sweet spot where one would prefer TfIdf over the modern dense vector representations (such as word2vec, doc2vec, etc.) ?
in cases where all the 3 documents contain the word, even if 2 of them contain the word only once and the 3rd doc contains it a 100 times, tf idf would be 0 as idf would be 0. isn't this misleading then?
Bro, you are a good narrator but a bad organizer. It would be better that the next time you write on the board more regularly in order to make it easier to follow what you sayin
This is really good. Concise , straight to the point, and there is no need to show a line of code !
I'm really glad to choose this video instead wasting my time watching 30minutes explanation of tf-idf. Great job for explaining this
Being a math lover, within a minute of your explanation I became your fan, was always in a search of videos like this
true. his channel hasn't been picked up by UA-cam yet.
I read this explanation in a book, but not as clear as this video. Well done!
thanks!
Your videos before sleep... Keep nightmares away...
aww, thanks !
Such a clear explanation!!! Much better than my teacher in the class. Why can't they just make it this simple? Thank you so much.
no problem!
always the best place to look for a concept explained. Always grateful.
I started googling tf-idf and then I was like "Hey, maybe that guy has a video on it", and you do! Thanks!
😂 "that guy" says you're welcome
@@ritvikmath haha sorry, Ritvik!
Thank you for this! You saved me much time! Your explanation is legit!
Thanks!!
Excellent teaching! Perfectly designed, clearly explained and not even one sentence that would be redundant. I’m your fan my friend 👍🏼🙏🏼
Wow, thank you!
@@ritvikmath Yes excellent explanation
Thank you for the video, we are working at a Movie recommender System and this helps a lot for NLP.
Your examples are excellent! Thank you!
You're very welcome!
I wish to have your coherence when explaining. Awesome explanation as always.
I like your videos first and then start watching your Data Science videos because I am sure that after I am done watching it, I will like it anyway.
Keep it up.. 🙏
Wow, thank you!
if the word 'healthcare' did occur in all 3 speeches, but occurs in the Obama speech 26 times, but only once in Clinton's and Bush's speeches. Using this mechanism, the IDF of healthcare would still be 0, but since the word has been used a considerably large number of times in the Obama speech, it is definitely important
in a more realistic situation the # of D would be much larger so cases like this would be extremely rare
Good point
Lucid explanation, my man back at it again!
When using the whiteboard, your videos are even better than with pen and paper! Thanks for your videos!
Cool! Loved your simple but extremely efficient explanation
What a classy explanation. So good man!
Much appreciated!
great video with depth and simplicity at the same time!
What if the word we are checking does not appear in any of the document, then in the denominator it would be 0 which is not possible
So simple and concise! Thank you so much!
Excellent , simply briliant
For any given word/term, we want to know how important is that term for a given document, relative to the entire corpus of documents. E.g. for Clinton these subset of words is really important in his inauguration speech, relative to the other inaguration speeches. TF-IDF is simply a multiplication of the metrics TF (term frequency) and IDF (inverse document frequency).
that explanation was so smooth and clear.. great job
Thank you :)
veyr great explanation, much better than my lecturer
Outstanding explanation!
Awesomeeee Simple and Clear
Great presentation!
If anyone dislikes this explanation god will have to come down to explain him/her.
Excellent explanation !
Thank you so much for explaining this clearly sir
Nice explanation. Thanks!
Very succinct explanation, thank you very much
You are welcome!
You saved me! My professor explained this in 3 hours, I watched it 2 times and I don't get it. This guy explained the same concept in 7 minutes and I get it!
clear cut explanation. Thank you
Youre an excellent explanar man. And I don't mean that lightly (I rarely compliment people wallah).
You got a knack. Truly!
Subscribed!!
I appreciate that!
Amazing explanation!
Good explanation in a simple way... keep doing well man
Thanks a ton!
How do you model multiple objects associated to a term class: Dental Care: United Health Care, Blue Shield, ..., by state? This becomes contextual and local within the text - how close is the word dental care in the text to UHC, for instance. The result would show which states address dental care in their health insurance regulations and which insurance companies make it available - both in a positive and negative way. Understand that this is a narrow example. Thanks
This came in clutch, thanks
Hi, first of all, thanks for the great explanation. I have watched your videos about Word2Vec and TF-IDF, and I need help, please. I'm a student working on a project about binary classification of SQL injection attacks. The dataset I have contains two columns: 'sentence' and 'label.' I need to extract features, but I'm confused about which technique to use: Word2Vec or TF-IDF. Can you help me decide?
Powerful...Thank yoiu
Explanation was awesome!
Great explanation buddy🙌🏻
That was crystal clear, thanks
Great explanation
Great Job sir!
Amazing stuff, thanks man for letting me pass the exam.
Perfectly explained
Would you advise to take out stopping words and run tdidf on the new set of documents?
Very well explained
Very useful! Thank you Sir!
YES! I get it now, much love bro
Is this a good tool to create a top of "important" words in a dataset? or it just helps to see the relevance in a particular document, I want to use it so I can maybe sum all the tdidf of all the documents and create a top words but I don't know if this is the best approach/solution to what I want, thank you in advance
Nice explanation!
Sweet and simple!
Superb!!
Awesome video!!
Thanks!
Clear and concise.
Excellent !
This is a great explanation. Thanks.
I have a question about differences between the implementation described in this video and another implementation commonly found on the web.
Can you explain how these two details would impact the final representation:
1) Term frequency simply calculated as term count
2) Applying vector normalisation (L2) to the document vector obtained in this video
Another question which is more open-ended: why is TfIdf still relevant ? Or less provocatively - is there a sweet spot where one would prefer TfIdf over the modern dense vector representations (such as word2vec, doc2vec, etc.) ?
Nice Explanation
Thanks!
iT IS REALLY NICE. KEEP IT UP
Thanks a lot 😊
damn.. that was a solid explanation
Great video! Thanks! I would love to see more content on TFIDF.
Noted!
I wonder why my teachers couldn't explain so simply.
love that in this alternative timeline the last speech is from Obama
a certain president would really bias the vocabulary data
Thank you so much!!! 🤩
Amazing!!!!!
thank you very much
in cases where all the 3 documents contain the word, even if 2 of them contain the word only once and the 3rd doc contains it a 100 times, tf idf would be 0 as idf would be 0. isn't this misleading then?
many thanks
very Good
Thanks
amazing
Thanks , great teacher if I could I would have given you 3 thumb
Useful :)
Glad you think so!
but if healthcare appears 100 times in one document, and only once in each of the other 2 documents, then the result will be zero!
This was my question. If you found out let me know.
Great video btw, best explanation.
It was that easy
Bro, you are a good narrator but a bad organizer. It would be better that the next time you write on the board more regularly in order to make it easier to follow what you sayin
Thanks for politicising education with that exclusion with that example, unsubbed - so partisan.
Sorry to see you go, it was not my intention to politicize but rather just to use this as an example.
That was a great explanation, Thanks 🤍
your explanations are great bro cut to the heart of the issue + ensure conceptual understanding 🫡🫡
Thank you so much 😀