Article classification

Objectives

This python script provides an easy topic separator - given a set of documents, and a number of topics, as well as two ready to use vizualisation tools.

Use

The class TopicSeparator is initialized with a set of documents. The function categorize() enables the classification by topic, given a number of topics. For each topic found, the user is asked to enter labels given the ten most defining words of the topic (the algorithm uses NMF with the tdfidf version of the documents):

The two functions plot_topics_split() and plot_words_importance() enable to visualize the categorisation:

plot_topics_split():

plot_words_importance():

This examples uses 800 articles from NYTimes, on the September, 10th 2014.