You are working for the data science team of Twitter that needs to classify people's tweets as y1= "hate-speech" or y0= "not-hate-speech". You have access to a dataset of thousands of tweets each represented as a sequence of words x and the true label y. A. (5 points) Assuming that all words came from an vocabulary of sizer V, suggest a way to map the i-th word of the tweet text into a vector xi. B. (15 points) Draw and dimension a CNN-based architecture that that will successfully predict accurately the label of the tweet yhat. Please note that all explanations of the choices you have done, such as dimensions of all tensors involved, must be shown to get full credit and its up to you to define the number of layers of the network. C. (10 points) The vast majority of the posts are "not hate speech" but a much smaller fraction are. Define a loss function that will address the imbalance of such binary classifier