Content Based Hierarchical URL Classification with Convolutional Neural Networks
Kishan Maladkar · 2019
Classifying the URL into respective user segments helps the marketing researchers to identify the intent and the preferences of a user. The combined words in the URL along with the punctuations makes it very hard and challenging to classify. The proposed model and the architecture in this paper addresses the problem statement. The taxonomy under which the URLs are segmented follows a hierarchical pattern. We follow an ensemble nature of classifying the textual content into the respective branches, depending on the unique confidence values for each branch. With the growth of the digital work on the internet day by day, assigning a single category for a particular class of text will yield less meaning compared to assigning a hierarchical class which also helps the marketing and advertisement industry. We perform the classification of the text into a pre-defined hierarchical taxonomy using an approach that uses a combination of multiple neural networks to derive the results.