Distributional semantics can represent also a larger chunk of text that consists of words,
such as phrases, sentences and documents. In this task, you will build document representations
based on doc2vec [2].
Doc2vec is an unsupervised algorithm to generate vectors for sentence/paragraphs/documents that represent a larger text in a low dimensional space similar to word embeddings. The algorithm is an adaptation of word2vec. Once you learn the document embeddings, you will train a classiffer to group a set of documents based on their topic.
You will train doc2vec model using Gensim3 in Python and train Logistic Regression classiffer using a library. You are free to use any library for the classiffer, but we recommend sckitlearn.