Customer Response Analysis Based on NLP of Sales Activity Texts

This study aims to develop an automatic customer response classification system using NLP (Natural Language Processing) techniques in a B2B sales environment. By utilizing unstructured text data such as meeting notes, emails, and chat records, the study aims to optimize sales strategies through accurate analysis of customer response. The proposed model is a deep learning-based architecture using BERT (Bidirectional Encoder Representations from Transformers) embeddings and an Attention mechanism within a BiLSTM (Bidirectional Long Short-Term Memory) model, classifying customer responses into positive, neutral, and negative categories. The model was trained on a total of 2,865 augmented text samples and achieved a validation accuracy of 89.35%, showing particularly high performance in classifying negative responses. Performance metrics such as Accuracy, Precision, Recall, and F1 Score were used, and the proposed model demonstrated overall superiority compared to existing models. Despite the limitations of domain specificity and response classification categories, the proposed approach shows significant potential for enhancing B2B sales decision-making. Future research should focus on expanding data diversity and integrating multi-modal data to enhance customer response analysis.