C3W4 Assignment Accuracy not reaching 80%

Course Q&A Natural Language Processing NLP with Classification and Vector Spaces
1

Hi all,

according to the suggestions in the notebook, I have tried adding a bidirectional LSTM, changed the number of units in it, however the accuracy just isn’t reaching 80% or more

Does anyone have any inputs on this?
Thanks

2

Hi mugdhagovilkar,

My sense is that in order to get higher accuracy you have to go beyond the structure of the model presented in the assignment. If this has your interest, you can have a look here or another source you may be able to find using bidirectional LSTMs in siamese networks to evaluate question duplicates.

3

With the following model you can get an accuracy of almost 85%

    model = Sequential()
    model.add(Embedding(total_words, 100, input_length=max_sequence_len-1))
    model.add(Bidirectional(LSTM(128)))
    model.add(Dense(total_words, activation='softmax'))

    model.compile(loss="categorical_crossentropy",
                  optimizer="adam",
                  metrics=['accuracy'])
4

Yes, this does work.

Looking at “C3_W4_Lab_2_irish_lyrics.ipynb”, it uses Adam(learning_rate=0.01). If this is used for this assignment, the accuracy can only reach at 0.7. Can I ask why 0.01 was used at “C3_W4_Lab_2_irish_lyrics.ipynb”?