Natural Language Processing Applied to Emotion Analysis for Continuous Improvement in Industrial Environments

The objective of this study was to analyze emotions expressed in natural language within Ecuadorian industrial environments using natural language processing (NLP) techniques. A sample of 150 anonymized texts was used, extracted from workplace climate surveys, satisfaction forms, operational reports, and digital suggestion boxes from companies in the manufacturing, food, logistics, health, education, and technology sectors. The texts were processed with an NLP pipeline using the spaCy library for cleaning, tokenization, and lemmatization, and semantically represented pre-trained models such as multilingual BERT and Emotion BERT. Supervised classification identified six emotions: joy, sadness, anger, fear, surprise, and neutrality. Statistical analysis revealed that the most frequent emotion was neutrality (34% of the total), followed by joy (21%), sadness (18%), anger (12%), fear (10%), and surprise (5%). At the sectoral level, logistics was the sector with the highest proportion of joy (28%), followed by food (24%) and manufacturing (12%). Manufacturing, on the other hand, had the highest proportions of sadness (20%) and anger (18%). The temporal analysis of positive sentiment between January and June 2025 showed a sustained increase in the technology sector, which rose from 62% to 77%. The healthcare sector showed an oscillating curve, falling from 58% to 56% in March and recovering to 63% in June. Education showed a cyclical pattern between 49% and 54%, related to its institutional calendar. In contrast, manufacturing remained stable but low, fluctuating between 40% and 42%. These results show that PLN allows us to capture structural emotional differences between sectors, and that these emotions can be correlated with internal dynamics of recognition, workload, or participation. It is concluded that automated emotional analysis is a useful tool for monitoring the organizational climate, anticipating psychosocial risks, and designing tailored intervention strategies. Its systematic application is recommended in organizations with digitized internal communication structures, especially in sectors with high turnover or emotional vulnerability. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.