Feedforward Neural Network $($FFNN$)$ and LSTM implementation: based on problem $2,$ study and complete the following two Python notebooks $($uploaded to Canvas$)$ which compare the FFNN and LSTM on an NER task. Each notebook is partially implemented, and you need to complete the parts marked with "TODO" to make the programs work.

P$3\_$Named$\_$Entity$\_$Recognition$\_$FFNNs$.$ipynb

P$3\_$Named$\_$Entity$\_$Recognition$\_$LSTMs$.$ipynb

Use a total of $5$ tags: PER $($person$),$ ORG $($organization$),$ LOC $($location$),$ MISC $($miscellaneous$),$ and O $($non$-$entity$)$ to label the sentences. For example, the correct tagging of "Steve Jobs founded Apple with Steve Wozniak." is $$PER$,$ PER, O$,$ ORG, O$,$ PER, PER$).$ Note that when consecutive words constitute a named entity, such as "Steve Jobs" in the previous example, they should both be tagged as PER.

Also, you will need to use the GPU running environment in Google Colab for this codina problem. You can chanae vour environment as follows: