Opinion mining is a classification task where you are provided with a phrase sentence/word and you are supposed to classify text into the following three classes: positive, negative, and neutral. However, it happens that neutral is never taken as a class this turn into a binary classification problem. For instance, 1-"Pervaz hey Janoon is a great movie." 2- "Pervaz hey Janoon is not a good move at all, 3- Pervaz hey Janoon is a movie........". These are three different short reviews of IMDB and each one of them conveys different opinions. E.g., the first sentence denotes a positive, the second one negative whereas, the third one is neutral opinions about the movie Pervaz hey Janoon. Here, a sentence is denoted by s and a fixed set of opinion classes C= {C1, C2, C3), and a prediction class c E C. Ultimately, a training set of n labeled sentence looks like this, such as: (s1,cs), (5), C3),....,(Sn, Cr) and output is a learned class. It is worthening here not all words of any sentence are meaningful i.e., convey the opinions. But at the other end we cannot ignore them as without context we cannot understand the value of its neighboring words. Considering the following scenario answers the question given below. [2+2+2+2+2+2+2+2 = 16 marks] a. What type of learning problem is this? b. What will be the training/testing phase if we solve this through a computer? c. Write down steps involved in implementing phase? d. What can be possible attributes or features that can be used to solve this opinion mining problem? e. What type of classification problem is this? f. Name few applications areas where this type of automatically opinion mining can be used? g. Which algorithm best classify the above problem and why? h. How we can avoid algorithms from overfitting during the training phase