Problem: A restaurant chain wants to see whether customers like their foods or not. They hire you as an NLP scientist to do this task. You crawled their website, collect and annotate all the reviews. Your task is to conduct the experiment for this text data.
a. How many classes would you identify for this food review and why? Please list down the classes name as your preferences.
b. You train two classifiers - (i) logistic regression/softmax classifier, and (ii) a two- layer feed forward neural network. Please mention differences between these two classifiers.
c. Now you are deciding to pick only one classifier for the deployment environment. How would you design your experimental setup to evaluate these two classifiers? How would you decide to pick which classifier for the deployment environment? Please mention the details reason behind your decision.