Vectoring text data of dictionaries' values from pickle file

Question

I'm new to NLP and trying to learn it by myself and I am doing classification.

I have a pickle file with some data like this,

{'food' : {'f1.txt', 'f2.txt', 'f3.txt', 'f4.txt'}, 'sports' : {'s1.txt', 's2.txt', 's3.txt', 's4.txt'}, 'politics' : {'p1.txt', 'p2.txt', 'p3.txt', 'p4.txt'}}

I need to extract these text file, divide them into training and testing data, vectorize the text data, classify it.

I have loaded the pickle file, here is the code

import pickle

topic_file = 'topic_dict.pickle'

with open(topic_file , 'rb') as x:
    topic_folders = pickle.load(x)

I am pretty stuck here, how can I divide them into train-test and vectorize it, I have no experience with this so any kind of help is appreciated.

Thanks in advance.