@@NicholasRenotte Hey, pls do a detailed video on Beautiful Soup. Also my request is pls make a video on other NLP usecases - text summarization, topic modeling, NER, text generation, Q&A with pretrained models (BERT, ROBERTA, GPT and so on)
brother I wanna say that your videos are very easy to understand and along the way to follow practically in our computer please make complete course on NLP and OPENCV And Machine learning because you are very good teacher I very like you god give you long life with a lot of happiness
amazing video! I mixed the content from your code with a csv file that I got from extracting tweets and now I can do a sentiment analysis on so many subjects! Thank you!!
Very good pal. You keep it simple and at high level. I would be happy to see those two negative votes to elaborate further why they are not satisfied though..
Awesome video! If its to the extend of your knowledge, a subsequent video about fine-tuning a pre-trained model to a particular dataset would be very interesting. So train, test and validate the accuracy of a model pre- and post-training. I have read some articles which showed accuracy of a model can be improved significantly when trained on the particular dataset.
Definitely, plus the models become a lot more practical when tuned to your use case. Definitely will have a series of tutorials or one mega tutorial coming on it soon @Michael!
@@NicholasRenotte did you make a tutorial about this issue already ? I couldn't like your video more,, it's simple, informative, short and just straight to the point :) Thank you
Hey Nik! I've been looking left and right for a guide on how to interpret the logits from BERT and thank god you've uploaded something that explains it well! I liked and subscribed. (Maybe you could refer me to a source, where you happen to learn it? - looking one for the FinBert variant) Also, your crash course on that scraping part is straight on point and I love that you executed each line from it to show you how it gets transformed. You are truly blessed with an out-of-this-world tutoring skill!
Thanks sooo much @Capt_Kaplan! I'm kinda just doing a lot of googling to learn it atm, there's heaps of examples of custom models in the HF model repo though, check this out: huggingface.co/ProsusAI/finbert
Hello Nicholas Reynolds, I am a college student, I like your video very much, I keep learning, you are my half teacher, thank you very much, look forward to more videos
Great vid! Not one of the more popular requests here but would be interested on a deep dive in building scrapers. It's kind of a niche topic since it's not the most exciting part of ML or Data Science but at least I get to throw my hat into the table. Haha. Looking forward to more vids =)
Love this video! You have an amazing skill in teaching others and make it look so simple. I'm subscribing you and I guess on my journey learning python I will watch a lot of your videos. Thanks a lot!
Fantastic explanation on all videos. just wondering whether you can do a video on sktime - installation, introduction, processes , etc. Hope to hear from you. cheerio
Do you currently have a video that covers how to do this? ->> "Can't be bothered building a model from scratch?". i would like to learn how to build or fine tune BERT. thanks
I have a problem: RuntimeError: Failed to import transformers.models.bert.modeling_bert because of the following error (look up to see its traceback): No module named 'torch.utils'
Awesome video! But, I was wondering if text preprocessing (removing punctuations, stopwords, lemmatization, etc) is needed before encoding the reviews? Appreciated if can get an answer
instead of using a lambda function for adding the review scores to the dataframe, would it be equally as good to use list comprehension to get a list of the scores then creating a new column out of that in the dataframe?
How would I fine tune this because right now im getting some inaccurate reviews. For some some reviews it is giving me 3 when it should be 4 stars or it is giving 5
Love mito but had them on a zoom call - it does not work with streamlit app. Perhaps it does not work with any app outside of a Jupyter notebook not sure 🤷♂️
I am stuck as i want to add an input which will be predicted. So basically i want to add an input ( a user would write a line for which the sentiment should be predicted) but i am unable to write the code for it. Can anyone please help me.
Hmm this is helpful for reviews but not for pos/neu/neg sentiment analysis, which is what I was actually trying to get some examples of regarding score interpretation.
Hey Nicholas Great Video!!! I was wondering if I could change the url in the code and give the class name of another website's comments other than YELP will it still work...
Hey Nicholas, is there any way of changing the bert sentiment score from 1-5 to a floating value between 0-1, similar to for example the vader sentiment score? Or do I need another bert sentiment model for this? Thanks in advance :)
Hello Nicholas, many thanks for the vidoe, how can one implement this code in tensorflow or what advantage pytorch has over tensorflow in Bert implementation?
Hi, I'm just learning NLP and it's totally new to me. Just want to know how to know how tokens / text can be passed for particular nlp pipeline? Thanks in advance
Is there a bert multingual up to date? This program you are using, uses old versions of numpy... between others. If there is an up to date version, it would be amazing to know, where to find it, thanks!
Another great tutorial thanks man Have one question in this video we are able to extract reviews from page 1 only so how to extract reviews from all the pages? Thanks once agian
Multi-langual and Dutch too... A double Heineken for u🍻 Maybe another vid of text classification for different classes/categories based on Bert ( Bert-base, multilangual cased) Not specific sentiment but categorizing sentences/text into subjects like cars, beers,soccer...
You got it, actually started work on a mega NLP tutorial. Zero-Shot Classification would be perfect for that!! Can't wait to drink that Heineken this weekend, still hoping Hamilton beats Verstappen though 😂.
@@NicholasRenotte Looking forward👍 Until this new season the F1 races were dominated by Mercedes. It is like running tensorflow with gpu versus tpu 😁 If Verstappen or other talented drivers got the same racing car as Hamilton then the competition would be more interesting .... But of course I hope Verstappen will win!
@@NicholasRenotte Yes the anomaly detetection algroritjm of Red Bull could be improved and I think Hamilton has not trained with the rigt AI Gym RL algorithm 😎
Thank you so much. Your video is so helpful. Is it possible to make another one about how to do sentiment analysis with BERTweet for covid -19 tweets? Thanks again.
Hi Nicholas, thank you for the video! It's really helpful for my work. Also, I have a question that is it a must to do text pre-processing by applying NLP model using deep learning? because it seems like you didn't apply any text preprocessing method for the raw text you've scraped.
For these models they're able to handle raw unprocessed input as they have a pre-processing pipeline built in @lowwwleechi. If i got sucky results I might add some additional pre-proc to improve the results!
Can you make one on mT5 or m2m100 for language translation models and how they can be trained from huggingface. Would be super helpful for techies across countries where content is limited.
These kind of content is what most budding ML enthusiasts need. Thanks for this
Anytime, so glad you enjoyed it @Sahil!
@@NicholasRenotte Hey, pls do a detailed video on Beautiful Soup. Also my request is pls make a video on other NLP usecases - text summarization, topic modeling, NER, text generation, Q&A with pretrained models (BERT, ROBERTA, GPT and so on)
I love how you take your time to explain each sentences and what they do! Absolutely love it!
Very helpful as a beginner!
Keep up
Thanks again for this tutorial Nick. My challenge is to 100% finish all your tutorials. 3hrs, 5hrs, 8hrs, 100hrs. all of it.
brother I wanna say that your videos are very easy to understand and along the way to follow practically in our computer please make complete course on NLP and OPENCV And Machine learning because you are very good teacher I very like you god give you long life with a lot of happiness
BERT with multilingual benefit is great advantage for analysis of text (especially with large-diverse dataset ) . This is enlightening ! Thank you Nik
Your channel is a treasure for Data Scientist.
amazing video! I mixed the content from your code with a csv file that I got from extracting tweets and now I can do a sentiment analysis on so many subjects! Thank you!!
hi im trying that as well, did you convert your file to a data frame like in the video?
This was a really helpful one. Thanks for the step by step and detailed explanation.!👍
Very good pal. You keep it simple and at high level.
I would be happy to see those two negative votes to elaborate further why they are not satisfied though..
Awesome video! If its to the extend of your knowledge, a subsequent video about fine-tuning a pre-trained model to a particular dataset would be very interesting. So train, test and validate the accuracy of a model pre- and post-training. I have read some articles which showed accuracy of a model can be improved significantly when trained on the particular dataset.
Definitely, plus the models become a lot more practical when tuned to your use case. Definitely will have a series of tutorials or one mega tutorial coming on it soon @Michael!
@@NicholasRenotte did you make a tutorial about this issue already ?
I couldn't like your video more,, it's simple, informative, short and just straight to the point :)
Thank you
Hey Nik!
I've been looking left and right for a guide on how to interpret the logits from BERT and thank god you've uploaded something that explains it well! I liked and subscribed.
(Maybe you could refer me to a source, where you happen to learn it? - looking one for the FinBert variant)
Also, your crash course on that scraping part is straight on point and I love that you executed each line from it to show you how it gets transformed.
You are truly blessed with an out-of-this-world tutoring skill!
Thanks sooo much @Capt_Kaplan! I'm kinda just doing a lot of googling to learn it atm, there's heaps of examples of custom models in the HF model repo though, check this out: huggingface.co/ProsusAI/finbert
You are a life saver man. i was stusked with my research. you saved me.. i understood the basics. Thanks
Hello Nicholas Reynolds, I am a college student, I like your video very much, I keep learning, you are my half teacher, thank you very much, look forward to more videos
Thanks so much! Stick around, plenty more mega courses to come!!
I love the way you teach. I'm going to do my final year project on sentimental recommendation system. This will help me a lot
Amazing video! Simple, structured, keep up the good work !!! P.S This helped me so much on my course lol
Awesome video, now I need to find a Brazilian Portuguese trained model asap to test it.
Thanks!
Great video. Would love to see a scraping tutorial. Especially one where you could pass credentials and log in first
Great vid! Not one of the more popular requests here but would be interested on a deep dive in building scrapers. It's kind of a niche topic since it's not the most exciting part of ML or Data Science but at least I get to throw my hat into the table. Haha. Looking forward to more vids =)
You got it @John, was planning on doing a live stream on it! But might put it into my Python for Data Science course that's coming!
You made it sound so simple! This is apt for beginners like me. Thank you ^^
Love this video! You have an amazing skill in teaching others and make it look so simple.
I'm subscribing you and I guess on my journey learning python I will watch a lot of your videos. Thanks a lot!
This is best content on bert .. loved the explanation ..
Yes, i think a bigger web scraping tutorial would be beneficial
Did he make a tutorial about it?
Excellent video explaining multiple concepts. Thank you for the great content!
I'm a new subscriber.., loving ur videos already... Cheers man all the best
Thanks so much @First Attempt, welcome to the fam!
Incredible work, love your explanations. Thank you!
Great job, Nicholas!
Awesome video, so neatly explained step by step --- thanks a lot
Awesome video. Simple and easy to follow. Thank you.
really loved your video. Would like to see if you can do same video but for stocks such as NVDA or TSLA. Thanks.👍
exactly what i was looking for thank you you made my day🥰🥰🥰
Awesome stuff man, helping me on my project
Great work bro..pls make a video on activity recognition using mediapipe
You got it!
Nice vid.
It's an application of BERT, does not quite add much new science... But good enough for some hands on with it.
Keep it up 👍
Yah, kept this one super high level. Can definitely get way further down into the weeds!
Love your videos bro 😉
Feel like lot to learn
Thanks a bunch @Dheeraj! Definitely always something new!
Great video! simple and quick to get good results thanks a lot!
An excellent tutorial. Thank you much for this.
Great content, great delivery .... overall great video
Fantastic explanation on all videos. just wondering whether you can do a video on sktime - installation, introduction, processes , etc. Hope to hear from you. cheerio
Haven't heard of it before, is that a timeseries lib for scikitlearn? Sounds awesome if so!
How to validate the output of the sentiment?
awesome explanation. however i have some questions, is text preprocessing included in the BERT model? and what is the thing with 'x[:512]' limitation?
Would love a video on fake news detection.
Thanks for teaching us!
Ooooh, yeah that'd be sick, I'll add it to the list!
@@NicholasRenotte Thank you
Have a nice day
@@heisenbergwhite5845 you too!
Good conceptual video. I was your 1k'th like
Sweet video, loved it!
Do you currently have a video that covers how to do this? ->> "Can't be bothered building a model from scratch?". i would like to learn how to build or fine tune BERT. thanks
Very good explanation and content
Thanks so much @Muzammel! So glad you enjoyed it.
What’s the logic behind setting the slice up util 512 when we were making our predictions?
Hi Nicholas , can you explain me how can I scrape rating of particular user from the webpage you mentioned
df["sentiment"] = df["review"].apply()
here in this function its giving me error related to tensor size
pls help me
I have a problem: RuntimeError: Failed to import transformers.models.bert.modeling_bert because of the following error (look up to see its traceback):
No module named 'torch.utils'
best video, very well explained
Awesome video! But, I was wondering if text preprocessing (removing punctuations, stopwords, lemmatization, etc) is needed before encoding the reviews? Appreciated if can get an answer
Yes, in the preprocessing we remove all the noise from the data.
instead of using a lambda function for adding the review scores to the dataframe, would it be equally as good to use list comprehension to get a list of the scores then creating a new column out of that in the dataframe?
@Nicholas Renotte can u help me or tell me how to apply word embeddings in BERT for text classification
can you explain the advantage and disadvantage of using pre-trained model in our data?
How would I fine tune this because right now im getting some inaccurate reviews. For some some reviews it is giving me 3 when it should be 4 stars or it is giving 5
Love mito but had them on a zoom call - it does not work with streamlit app. Perhaps it does not work with any app outside of a Jupyter notebook not sure 🤷♂️
I am stuck as i want to add an input which will be predicted. So basically i want to add an input ( a user would write a line for which the sentiment should be predicted) but i am unable to write the code for it. Can anyone please help me.
Hey! When I change the website to another one (like Amazon), it doesn't go with it. Please help me with it
Again best explanation
Ayyye, thanks so much @Ameer!
Hmm this is helpful for reviews but not for pos/neu/neg sentiment analysis, which is what I was actually trying to get some examples of regarding score interpretation.
Thank you for the video. Can I run a content analysis on short Customer Feedbacks that's releable?
how do i apply by using a dataset i collect from kaggle? like excel dataset. Do i just change the link or it has to be from YELP
scraping a web page and using that data as training data within a chat bot would be very interesting.
Thank you for making a video on this
Anytime! So glad you liked it!
Amazing video!
May I know if this support a multi-language in one sentences or in one review?
Appreciate your work sir
Thanks for checking it out man!!
Hey Nicholas Great Video!!! I was wondering if I could change the url in the code and give the class name of another website's comments other than YELP will it still work...
Can we add fully connected layer on top of that model to use it for other text preprocessing tasks 🤔
Could probably pass through the logits to a subsequent layer!
Awesome as always!! I really appreciated this. Did you finally made a web scraping for data science super tutorial!??. Please tell me you did!😃
can you tell me how to save this model and use it
Hey Nicholas, is there any way of changing the bert sentiment score from 1-5 to a floating value between 0-1, similar to for example the vader sentiment score? Or do I need another bert sentiment model for this? Thanks in advance :)
Need another BERT model @Michael, this one is specifically trained to give you a binary value.
Please make a video tutorial on Web scraping as well.
Great tutorial thanks
Is there any method to evaluate output ?
Hello Nicholas, many thanks for the vidoe, how can one implement this code in tensorflow or what advantage pytorch has over tensorflow in Bert implementation?
How can you make a sentiment analysis on languages other than English ?
Hi, I'm just learning NLP and it's totally new to me. Just want to know how to know how tokens / text can be passed for particular nlp pipeline? Thanks in advance
hey what a great videos, can you make a tutorial for sentiment analysis deployment to a simple web?
Muy util. muchas gracias por la informacion. Saludos desde Mexico
Excellent Video.
Can you point me to the changes to be made if TensorFlow is used?
Is there a bert multingual up to date? This program you are using, uses old versions of numpy... between others. If there is an up to date version, it would be amazing to know, where to find it, thanks!
if not, could you especify the version of nunmpy that was used
I am interested in Data science web scraping. How do I go about it?
the video is good, very petrified for ordinary people like me. but how to test the accuracy of the bert that is used?
Can you do sentiment analysis using apache spark - pyspark mllib
Hello Nicholas!, i got a question is there any ML that can extract keywords from a text?
Could use Named Entity Recognition to extract keyword types. If you're just looking for specific keywords you could just use a find function!
@@NicholasRenotte Thank you, love your work!
You’re the best around!! Thank you for this!!
One question: is it possible to extract comments from a Facebook post?
Thank you.
I think it's pretty tricky, FB is notoriously hard to scrape. Can probably get your own comments via the API?
Another great tutorial thanks man
Have one question in this video we are able to extract reviews from page 1 only so how to extract reviews from all the pages?
Thanks once agian
Could loop through the subsequent pages as well @Nandan! Would need to build a more sophisticated scraper for that however.
Cool and informative.
bert is too big (1GB for the one with good quality). What would be the best way to deploy it with my django back end? Any recommendations?
Deploy it as an API or use a pre-hosted instance (they're available direct through HF Hosted)
how can we use it on other sites like bbc news?
Can I use BS4 to web scrape Twitter?
Multi-langual and Dutch too... A double Heineken for u🍻
Maybe another vid of text classification for different classes/categories based on Bert ( Bert-base, multilangual cased) Not specific sentiment but categorizing sentences/text into subjects like cars, beers,soccer...
You got it, actually started work on a mega NLP tutorial. Zero-Shot Classification would be perfect for that!! Can't wait to drink that Heineken this weekend, still hoping Hamilton beats Verstappen though 😂.
@@NicholasRenotte Looking forward👍
Until this new season the F1 races were dominated by Mercedes. It is like running tensorflow with gpu versus tpu 😁 If Verstappen or other talented drivers got the same racing car as Hamilton then the competition would be more interesting .... But of course I hope Verstappen will win!
@@henkhbit5748 you almost had your wish this weekend 😂 if those tyres didn't blow Verstappen was on a dream run!
@@NicholasRenotte Yes the anomaly detetection algroritjm of Red Bull could be improved and I think Hamilton has not trained with the rigt AI Gym RL algorithm 😎
@@henkhbit5748 😂 😂 😂
Thank you so much !
Thank you so much. Your video is so helpful. Is it possible to make another one about how to do sentiment analysis with BERTweet for covid -19 tweets? Thanks again.
You got it! Working on a bunch of NLP stuff atm.
@@NicholasRenotte Did you make a video about BERTWEET? I didn't find it.
@@ey5722 nope, not yet.
Hi Nicholas, thank you for the video! It's really helpful for my work. Also, I have a question that is it a must to do text pre-processing by applying NLP model using deep learning? because it seems like you didn't apply any text preprocessing method for the raw text you've scraped.
For these models they're able to handle raw unprocessed input as they have a pre-processing pipeline built in @lowwwleechi. If i got sucky results I might add some additional pre-proc to improve the results!
@@NicholasRenotte Got it! Thanks a lot!
Can you make one on mT5 or m2m100 for language translation models and how they can be trained from huggingface. Would be super helpful for techies across countries where content is limited.
You got it! Got a ton more HF stuff coming as well.
does these code works for multilingual comments ?
Hello, instead of doing the first 512 tokens, aren't you just taking the first 512 characters? Lov your videos btw :)
Can BERT be used on time series data I wondered ?
It's mainly for NLP :), take a look at the Facebook prophet vids for time series forecasting!