ML Classification using GloVe Vectors & Keras ❌NLP Project in Python with GloVe, TensorFlow & Keras

[@lukasga3729]

Thanks for the tutorial and that you show the code step-by-step with result, rather than just discussing the code

[@DecisionForest]

Thank you, I believe this is the best way to explain concepts, seeing the intermediary results

[@ijeffking]

Excellent tutorial with a good approach to teaching what each block of code does. Thank you so much. One of your best tutorials so far...

[@DecisionForest]

Thank you, I'm thinking to switch the teaching method a bit to make it more dynamic. Once I create new tutorials would really appreciate your input on those, see how they compare.

[@ijeffking]

@@DecisionForest Sure. Waiting for the new ones...

[@anonymousa7581]

Your video is gold ❤

[@DecisionForest]

Thank you!

[@dans722]

nice vid , why was 50 chosen as the max length and does it have an effect on the model? thanks

[@ihiy9908]

Im sorry im quiet confused if what i want is the word vector model that made by glove from my own corpus, how can i get that? So i can used the word vector leter to classifying another data

[@bobbygia1198]

Is the num_words used correctly? You are just using the size of the training set. Shouldn't the num_words be the number of tokens that appear in the training set?

[@priyadarshiroy2455]

I am getting similar results for a nlp project I am working on but the difference in accuracy between the training and validation data is similar to yours. There is about a 10% difference between them.....and like yours the validation loss does not decrease. Does this mean the model is overfitting? I am very confused. Can someone please help me....

[@precious_orim]

Hello pls were u able to figure this out ?

[@randb9378]

How do you suggest dealing with a word that is not in our vocabulary?

[@dc_8616]

Great tutorial.

[@DecisionForest]

Cheers Darren!

[@randb9378]

you are great at explaining! thanks

[@DecisionForest]

Thank you!

[@sadikaljarif9635]

is it possible to make lstm model with glove for fake news detection???

[@precious_orim]

Thank you sir. Pls u didn't put the link to the other video

[@zimin7255]

thanks for the great content, subscribed!

[@DecisionForest]

Thanks Zimin! Glad to have your support!

[@ajitkumar2670]

any idea how to determine max_len because i am having text of size more than 1000 words

[@DecisionForest]

hope I understand correctly but you can easily get the max value between the lengths of each record in the series.

[@ajitkumar2670]

@@DecisionForest Yes, I can get that. But I am confused that in various videos suppose if the max length of the sentence is 300 but people are using 50 as maxlen for padding.

[@ajitkumar2670]

@@DecisionForest and I have confusion regarding vocab_size also which the number of unique tokens in the corpus. Is it always necessary to take that number of tokens or we can take less? Is there is any thumb rule for deciding the number of tokens that can be used for creating the matrix?

[@DecisionForest]

You should set the max length to the actual max length not less. To the question regarding the vocabulary size when you create the corpus it will contain all the words in the document. You need to use the whole corpus.

[@ajitkumar2670]

@@DecisionForest There are some words whose frequency is only 1 based on the tokens generated by the tokenizer so is it good to drop those words or we can keep that.

[@sallahamine9467]

Thanks for the tutorial -> model.fit(train_padded,train_labels, validation_data=(test_padded, test_labels), epochs=20, verbose=1) Error:- UnimplementedError: Graph execution error: How to fix this error? 😥

[@humairaafia5658]

Thank u soo much ,dear. Please, make a tutorial on "Character Embedding" for text classification , if possible.

[@DecisionForest]

Glad it was useful. Very good suggestion, do you have a use case where Character embeddings would be more suitable than word embeddings? As there is less semantic information available there even if they work well out of sample.

[@humairaafia5658]

@@DecisionForest There is less work done using character embedding in my language(non-english). So i want to apply it and see the impact. And it would be very kind if u make a tutorial on it and help me understanding this topic. Thank you.

[@LRth3KING]

Hi nice tutorial, can the Glove Vectors support portuguese words? Thanks you you helped me a lot

[@DecisionForest]

Glad you found it helpful! From my knowledge it only supports English but there are other options for pre-trained word vectors in other languages for sure.

[@LRth3KING]

@@DecisionForest Can you help me with something else? I wanted to use the Glove model to use for input in a Logistic Regression model. But I don't know what variable should I use from here

[@DecisionForest]

@@LRth3KING Don't understand what you mean by which variable to use but does this help? kzbin.info/www/bejne/opqylounps2grcU

[@LRth3KING]

@@DecisionForest by variable I mean the X_train and X_test values for the Log. Regression model. I already use the Tf-if Transformer and countvectorizer. I want to test various inputs and see who performs best

[@DecisionForest]

Oh yes, you want to add more features to the train and test sets. Well you can add any new features that you create, such as word length and others alongside the word embeddings and then train the model.

[@jkally123]

would be much nicer and more helpful if you mentioned where to take the data

[@DecisionForest]

Hi Yossi, at the beginning on the video at 00:37 you have the link to the dataset in the notebook.

[@aanchalmahajan3821]

Please share the file.

[@adrienloridan]

Too many ads

[@subashchandrapakhrin3537]

Your Validation loss is increasing. Good Video Though, I loved it.

[@DecisionForest]

Cheers mate

[@sarabassir3032]

hello bro can you send me your email thank you in advance

[@aayamoza8495]

Awesome tutorial.