Text analytics to improve telecommunication customers service management from unstructured data

Mariana Esteves Oliva

Key information

Authors:

Mariana Esteves Oliva (Mariana Esteves Oliva)

Supervisors:

Paulo José de Jesus Soares (Paulo José de Jesus Soares); Maria da Conceição Esperança Amado (Maria da Conceição Esperança Amado)

Published in

January 27, 2021

Abstract

The digital transformation is arriving at a fast pace into all areas of business. Companies feel the need to change their internal processes and automate tasks to keep up with the digital transformation happening around the globe, changing the way they deliver value to the customers. This research explores the current state of the art technologies, applied to the telecommunication call center industry, such as big data analytics, text mining, NLP and machine learning models. These technologies will be applied to troubleshoot telecommunication technical problems in service call centers, providing alternatives to enhance the troubleshooting process. This will be achieved through the development of a recommender system which will recommend the most probable causes of being chosen by the assistant regarding a problem reported by some customer. This recommender system uses as input the descriptions of the calls in the form of text. Several NLP techniques are applied to the text during the pre-processing phase and three text classification models are proposed: a Naive Bayes model, a Random Forest model and a Recurrent Neural Network. Ending this research, the three models are evaluated and compared, resulting in an out performance of the Random Forest model both in accuracy and training time.