Uso de modelo de linguagem LLM para análise de sentimento e resumo de ligações em call center

Abstract

The present work proposes an Artificial Intelligence solution for analyzing phone calls in a call center of an electric utility company, using exclusively Google Cloud Platform (GCP) resources. The motivation arises from the need to optimize processes, reduce costs, and improve customer relationships in a scenario of high demand for support and complaints. To this end, techniques of Natural Language Processing (NLP) and Large Language Models (LLMs) are explored, focusing on audio-to-text conversion, sentiment classification, and automatic summarization of interactions. The audio data is initially stored in Cloud Storage; then, it is transcribed into text and imported into BigQuery, where BigQuery ML (BQML) and Vertex AI resources are employed for the development and integration of analytical models. The proposed approach allows for identifying the main issue, assessing customer sentiment (positive, negative, or neutral), and summarizing the content concisely. In this way, the company can prioritize critical cases, monitor customer satisfaction, and obtain valuable insights for continuous service improvement. The results confirm the effectiveness of applying LLMs to support decision-making in call centers, demonstrating that GCP cloud solutions offer scalability, reliability, and security to handle large volumes of data efficiently.