An Analysis of Machine Learning Techniques to Prioritize Customer Service Through Social Networks


  • Paulo Roberto Pessoa Amora Federal University of Ceara
  • Elvis Marques Teixeira Federal University of Ceara
  • Maria Isabel Vasconcelos Lima Federal University of Ceara
  • Gabriel Maia Amaral Federal University of Ceara
  • Jose Ricardo Anacleto Cardozo Digitro Tecnologia SA
  • Javam de Castro Machado Federal University of Ceara


Sentiment Analysis, Deep Learning, LSTM, Social Networks, Customer Service


The large amount of opinionated data made available by social networks allows the extraction of valuable information for a variety of applications. Sentiment analysis is a powerful tool in this sense, allowing to identify and classify opinions in texts according to the predominant polarity exposed in them. An interesting use of this technique is for companies to rank the messages from their clients in order to identify and attend the most dissatisfied ones first, thus improving customer service. In this work, we evaluate the application of a range of different machine learning techniques (including two deep learning ones) to the sentiment analysis of tweets in Brazilian Portuguese, aiming customer service prioritization. Our results show that the deep learning models are able to classify tweets more efficiently in this context, compared to traditional machine learning ones.


Download data is not yet available.