Referring to the example given in the keras docs here: https://github.com/fchollet/keras/blob/master/examples/imdb_bidirectional_lstm.py
I would like to use my own dataset instead of IMDB. After inspecting the format of the default dataset, i see that each word in the sentence is replaced by its vocabulary index, which is sorted in descending order.
I was looking through the keras docs here https://keras.io/preprocessing/text/ for a method that would accomplish this, none of them seem to work for me.
I have been trying the
Tokenizer.fit_on_texts and Tokenizer.fit_on_sequences methods.
Fit on texts returns a
AttributeError: 'float' object has no attribute 'lower'
error.
My input is a pandas series of text.
Could anyone point me as to what I'm doing wrong? I have looked at the following thread and it did not help
Keras - Text Classification - LSTM - How to input text?
Thank you!
