Web app will take article abstract which is not added yet in the system as input by the user and on clicking generate recommendations your application will infer user article vector. Then it shows articles recommendations by computing cosine similarity between user vector and already added article vectors in descending order.