Training an AI to create poetry (NLP Zero to Hero - Part 6)

Through this series so far you’ve been learning the basics of NLP using TensorFlow. You saw how to tokenize and then sequence text, preparing it to train neural networks. You saw how sentiment in text can be represented with embeddings, and how the semantics of text over long stretches might be learned using recurrent neural networks and LSTMs. In this video we’ll put all of that together into a fun scenario -- creating a model and training it on the lyrics to traditional Irish songs.
Irish songs generator Colab → goo.gle/3aSTLGx
Predict Shakespeare with Cloud TPUs and Keras → goo.gle/2zy4A40
NLP Zero to Hero playlist → goo.gle/nlp-z2h
Subscribe to the TensorFlow channel → goo.gle/TensorFlow

[@Lord-V15]

BEST SERIES ON THIS CHANNEL. SAD THAT IT'S OVER :(

[@MrChughad]

Train a model with these 6 videos, then you can generate further videos ;)

[@ShermanSitter]

I've watched this 3 times and understand it more each time. You are a great teacher and thank you for the content!

[@rodrigoriquelme527]

I've been looking for weeks material to introduce into NLP and this is by far the clearest lessons I've found, even being published two years ago, great teacher.

[@edwinbuabeng3847]

Very simple explanation. Made everything look very easy. Thank you so much for making these. I would like to see more videos in this series on NLP

[@asadanees781]

Hi Laurence, Thanks for recording NLP series and amazing . This is blessing that it is freely available. It would be great to learn much more recording about NLP.

[@sarfarazansari9053]

Amazing series. Thanks a lot for creating this series....You have explained it so well that things start looking simple. It creates lot's of interest and confidence. Proudly to say "I have leaned the deep learning and NLP concept by watching Laurence Moroney training video'"... Looking forward to have more training videos like this from you.

[@Powell___]

It is amazing how you able to explain, thank you for that. It would be great to see and learn much more regarding to NLP

[@LaurenceMoroney]

Thanks!

[@rezajafari5348]

Best Tensorflow learning Ever I've seen

[@mans9318]

This vid totally made me shed love raw boo...:p

[@christophereng9801]

Awesome video. Would love more content around topic modeling/generation!

[@stevemurch3245]

Really terrific, Laurence, thanks so much for all the effort you put into these excellent videos.

[@dethswurl117]

These videos are absolutely amazing. Thank you so much for making these, you're really helping newcomers to the community like me :)

[@laurencemoroney655]

Welcome!

[@MohammadYs77]

Best NLP Series I've ever seen. Thank you

[@adamcohenhillel]

I would like to see more videos in this series on NLP! You are a great instructor, Thank you!

[@visheshkapoor8466]

yo I found the optimizer function

[@0xsercaksoy]

playlist was amazing, thanks for work!!

[@soranlanguage8302]

Your method of explanation is amazing I sat hours for tensorflow tutorials, now I listen to you, I understand clearly and learn more. I hope you can do more video content on NLP

[@laurencemoroney655]

Thanks Soran! :)

[@mohitjoshi8818]

This was so helpful, thanks for this. It would be great if you make more like this.

[@xintel2001]

Currently, I am on Week 1 of the TensorFlow in Practice course. These nuggets of learning are a great way to impart knowledge on NLP and complement the course material. The notebooks are useful too. Thanks for your effort.

[@LaurenceMoroney]

My pleasure! Enjoy :)

[@syedadnan4295]

Really awesome to learn about NLP in such a easy way

[@525sumithmeesa9]

I am amazed that how can you explain such a detailed content in just 6 videos.....YOU ARE A LEGEND SIR Plz explain about voice recognition using tensorflow

[@luis96xd]

This was an amazing video! Thanks I want to see more NLP videos 😄

[@karanbrokendream]

Thank you for these videos.. the simplest and best explanation i've seen by far.

[@user-vj9hb3gy6d]

Wow, I can finally understand how a language model like GPT-3 can be made! I know that the GPT-3's model architecture is much more complex than this, but I can at least imagine how could that be achieved. Up until now, I looked at GPT-3 like a stone age man looked at fire. Thank you so much Laurence Moroney! :)

[@gabrielchung6325]

Very nice series. It gives me good understanding of NLP and I would like to thank you. More please.

[@laurencemoroney655]

Welcome!

[@DaveOmri]

thank you for this serie, fun till the end

[@omidforoqi4163]

OMG, very nice and thank you google. I hope this course continues

[@markusobi4148]

Very interesting series, thank you!

[@quadraticlife8314]

please create more videos, they're great, thanks a bunch.

[@26MS10]

Damn that was so easy to grasp, thank you so much Sir!!!!!

[@therohans5492]

This was a great series thank you

[@alejandromartinherve2534]

Is sad that this is over. I personally will love to see more on NLP combined with new novel approaches like Attention and BERT. I ❤ TF.

[@rohanmanchanda5250]

🎶 You just want attention. 🎶 🎶 You don't want their heart. 🎶 🎶Maybe you just hate the thought of them doing something new. 🎶 🎶 Yeah you just want attention 🎶 🎶 I knew from the start 🎶 🎶 Can't think of more lyrics that'll suit the case of you 🎶 Waaaaaaaaaaaa 🎶

[@danialaflakian2748]

Thank you for your beautiful explanation.

[@ayca7401]

thank you for the series!

[@kenjopac4247]

I literally had a project about this exact exact thing, Thank you very much!

[@hebagamaleldin8803]

It's simple and powerful, please provide us a series as simple as this in reinforcement learning and also deeper level of detail in NLP, Computer vision. Thanks so much.

[@laurencemoroney655]

There's already a computer vision one out there. I'm looking into RL

[@Ab_Jannadi]

It is a great series.. thanks , good work

[@leonardvanduuren8708]

Thank you so much Sir! Please give us the opportunity to learn more from you.

[@kimjensen1787]

Brilliant. Very vel explained. thanks

[@e_hossam96]

wonderful and simple. thanx.

[@esooghazy]

Thanks for the videos! It would be great if you can create more videos about NLP like creating custom spelling correction and text matching.

[@LeonardoPinto]

great content, we need more!

[@LaurenceMoroney]

Working on it...

[@jrodthriller]

great stuff, i'm getting a lot out of the series. thanks!

[@LaurenceMoroney]

Thanks! :)

[@sebastianandrescajasordone8501]

This was great!

[@emadgirgis6689]

Awesome work! Thank you. Could you please add videos about using transformers and attention?

[@adnan_randomnumber]

More videos please, the instructor is very good

[@shardsmp8831]

new content about nlp please, this series is amazing!

[@hikmetcancelik]

Yes we want more please!

[@b1twis3]

Great content! Thanks a lot

[@reeshmakannampurath5592]

Excellent!, Looking forward to Text summarization.

[@LaurenceMoroney]

Great idea -- I should work on that

[@cr0wzzz]

I've watched this 3 times and understand shed love raw boo more each time.

[@ghousethanedar7852]

Great content.....Thank you very much would like to see more video

[@theotherstuff2112]

Would love more NLP stuff!

[@laurencemoroney655]

Do you have any specific requests?

[@soufianei4941]

thank u ser , this series was very helpfull

[@humnashafqat9719]

great and easy to learn thanks

[@prasannadas8418]

Amazing tutorial ,Thanks

[@laurencemoroney655]

Welcome! :)

[@mansoural-harbi6096]

Awesome >>> I will vote for another session on NLP

[@LaurenceMoroney]

Thanks!

[@alfaalkaaf5335]

wow this is what i need. thats literally takes a loop to generate the text at the end. i understand that why you can use "" token to stop generating the text instead limiting the result to 20.

[@nirajkumarsingh5478]

nice content..easy to understand.

[@rlb5261]

It is just amazing!

[@rayyanamir8560]

This is the best video on next word prediction (text prediction). Thanks a lot Mr. Laurence. :) #NextWordPrediction #TextPrediction #NLP

[@rishikambhampati2862]

Thanks a lot. Teaching at it's best

[@hadizand3753]

Excellent.

[@vietdinh-van8018]

Your idea is great.

[@LaurenceMoroney]

Thanks!

[@koponstanley4]

shed love raw boo = you guys are amazing

[@alelignmekonen3759]

OMG, I have no word to describe how greate your presentation is. What can I say? I liked your lecture like my favorite food. Can u add more video: Gated Recurrent Unit(GRU)?

[@topekekere6614]

nice presentation. I want more video content on analyzing tweets

[@hahustatisticalconsultancy8869]

very very helpful and thanks

[@Dmitriuso]

Hey Laurence! Thanks a lot for this series, you are absolutely great! I wonder if you could make some episode(s) on chatbots and different strategies to approach this part of NLP. I personally carried out some experiments with chatbots, using intents-based approach - when you basically make intents for different kinds of questions and provide the chatbot with answers for this particular case. Google Dialogflow is great for that, by the way. Still, it's more like a stimulus-reaction. Many people try to use Cornell corpus of movies scripts to train their models - I did, too, but wasn't too impressed with the result. I wonder if more intelligent and creative text generation would be possible for a chatbot.

[@laurencemoroney655]

I can look into it, but don't really work a lot with chatbots

[@darrenaldrich6185]

Nice video! Thank you :D

[@charanbirdi]

great explanation and thank you for all the efforts ~~, any chance if you could include advance NLP videos i.e. transformer, LLM?

[@abhisheke223]

Very nicely explained. It would be nice if some content on the encoder-decoder model with attention and transformer model for language understanding is made because I find very difficult to learn these by myself.

[@LaurenceMoroney]

I'll look into it! :)

[@songslover6435]

Dear Sir! You are the proficient teacher in your teaching and sharing knowledge for the learners all around the world. By the way, Can I get more videos that are related to LSTM? Please Sir.

[@herstar9510]

It would be great if the playlist was listed in order of what we should watch.

[@adityadhekane]

Great series , I would like to see topicmodeling using tensorflow

[@adityadhekane]

@@GoogleDevelopers Scenario - Maybe an extension of LDA using TensorFlow, word2vec can definitely be considered but looking out for a more intuitive solution with n-grams and an option to auto label the topics which otherwise need manual expertise. (Hope this helps if you want more details we can take it offline over an email)

[@roshanshetty3531]

Felt sad hearing the series end. Please continue for machine translation, attention model, bert.

[@LaurenceMoroney]

Will go back to do some BERT stuff at some point in the future, but need to teach the basics first

[@kamilfilipek2879]

I came back to your videos after a while and I got amazed again! I wish there was a wideo (or tf function) instructing how to convert predictions score (probs) into classes. For now, we have to play with code little bit :) Or I missed something...?

[@laurencemoroney655]

argmax?

[@fallguyjames]

predicted = model.predict(token_list, verbose=0) predicted = np.argmax(predicted) Change the predict_classes to predict and find the maximum (best) class from that returned array of word indexes.

[@mithunchandrasaha403]

Nice Explation.Sir.I want more

[@yashpatil375]

I would like to learn how to build an AI chatbot using NLP, which could understand what we are providing and reply accordingly. You really made it very easy to understand the basic concepts behind the scenes. By The Way, I am currently enrolled in the DeepLearning.AI Specialization course for Deep Learning and I would soon enroll in your course on NLP too. Deep Learning is real fun in the computer science field as in my opinion. Thank You So Much, Lots of Love From India😊🤗

[@aungzinpyae8606]

Super simple and super amazing tutorial! Please make a simple tutorial for seqtoseq machine translation model.

[@RSUtsha]

Did I comment already? I just wanna say, this was awesome..

[@nisarahamad1455]

You are awesome Laurence... I am very excited to learn from you. We want to learn more from you. Can you please explain how can we use RNN/LSTM in the real time business scenarios like Time Series Forecast modelling please. Thanks in advance...

[@laurencemoroney655]

Will work on a time series course soon!

[@nisarahamad1455]

Hello @@laurencemoroney655 hope you are doing well :) I am waiting for your Time Series course

[@topekekere6614]

which software was used to curate this video? Excellent video presentation

[@agentstona]

Laurence morony please come back and make more tutorials you are the best ..

[@rahuldeepattri9244]

More please!

[@LaurenceMoroney]

Pausing for a little while as I'm working on my next coursera courses

[@amilamad]

Although the series come to an end The great man will continue to send All the things man try to amend People must accept the errand

[@LaurenceMoroney]

Ha! That's hilarious :)

[@pa1nk1ller14]

Just awesome as always :) Would love to see something related to transformers

[@LaurenceMoroney]

Robots in disguise? ;)

[@pa1nk1ller14]

@@LaurenceMoroney hahahahaa just like that! an Optimus Prime on steroids

[@asimpoptani316]

Def want more :)

[@laurencemoroney655]

Thanks! :)

[@16876]

great, thanks

[@laurencemoroney655]

Welcome!

[@MuhammadAli-fn5kv]

this is waht i am loking for

[@laurencemoroney655]

Great! :)

[@olamilekanbanjo5609]

Please can you make a series on speech recognition?

[@fahemhamou6170]

تحياتي الخالصة شكرا جزيلا

[@DrJanpha]

Great

[@deepanshu.770]

Please make more education video on NLP

[@kiruthihannagarajan3221]

It's amazing and I could implement it in various string dataset. But I have a doubt. How to improve it to learn from multiple poems?

[@LaurenceMoroney]

It's already learning from multiple poems, I aggregated them all into a single file

[@SD-nn7qq]

I'm getting a pretty good accuracy, but then this message is coming in the editor: Also, no text is being generated. Does anyone know the solution?

[@phamthienank14hcm30]

hi, i'm stucking at the like index ==predicted that said i have to use a.all or a.any, how can i fix that ? it keeps generating 1 word only :(

[@saeeduchiha5537]

is this series really meant for someone with a beginner level in NLP?? I can understand most of the concepts but I don't know any of the (keras) functions used!

[@soronprfbs17]

Please more NLP!

[@laurencemoroney655]

Do you have any specific requests?

[@tooljerk666]

Maybe I don't understand, but the [0] in token list is just trying to limit the model to the first line of the first song, right? If we delete [0] we would have the model learn the whole corpus?

[@rajnish747]

WE WANT MORE NLP

[@akashraut3581]

I have searched everywhere and couldn't find a solution. How can we use tf.keras.layers.masking to skip time steps when training with sequences of images? I'm trying to implement a deep RL agent with cnn + stateful lstm policy using functional api. So in order to make the variable sequence data into batches of fixed size I need to use masking layer which works amazing with padded data which is 1d and not 2d like images. But I m not sure if it works with images . Can anyone help? I'll be very thankful