Supervisor: Gian Antonio Susto
Creation Date: 05/10/2025 15:52
This thesis focuses on the design and development of a personalized AI chatbot aimed at supporting customer service and after-sales activities. The project is carried out in collaboration with Statwolf S.r.l., (https://www.statwolf.com/) a company specialized in artificial intelligence, data analytics, and industrial digitalization solutions.
The main goal is to create a smart conversational system capable of interacting naturally with customers, understanding their requests, and providing relevant, context-aware responses. The chatbot will leverage machine learning, natural language processing (NLP), and knowledge management techniques to offer customized assistance throughout the entire customer journey — from product inquiries to after-sales support and maintenance.
Dataset type: Already acquired data
Dataset description: The data used in this project originate from customer service and after-sales environments provided by Statwolf and its partner companies. These data reflect real interactions between customers, service technicians, and support teams, and are essential for training and validating the chatbot’s Natural Language Understanding (NLU) and dialogue management modules. All datasets are anonymized and processed according to GDPR and internal data-handling guidelines to ensure privacy and confidentiality.
List of Methods: The methodology adopted in this project integrates data-driven artificial intelligence techniques with knowledge-based reasoning to develop a personalized chatbot for customer service and after-sales support. The approach combines several interconnected components, each addressing a specific aspect of the system’s functionality. The process begins with data preparation and preprocessing, where customer interactions, technical documents, and service reports are collected, cleaned, anonymized, and structured into a unified dataset. Text normalization, intent labeling, and entity annotation ensure that the data are suitable for natural language model training and semantic understanding. The Natural Language Understanding (NLU) module forms the core of the chatbot’s intelligence. It employs modern transformer-based models to detect user intent and extract key entities such as product names, serial numbers, or error codes. This enables the chatbot to interpret technical language and contextual information accurately. Dialogue management governs the logical flow of the conversation. A hybrid approach—combining rule-based control for structured processes and machine-learning models for adaptive behavior—ensures both reliability and flexibility. The dialogue manager maintains conversational context, user state, and task progression across multiple turns. For response generation, the system integrates retrieval-based and generative AI methods. Semantic search techniques retrieve relevant information from the knowledge base, while generative models provide natural-sounding responses when predefined answers are unavailable. A ranking mechanism selects the most contextually appropriate output. Personalization is achieved through user profiling and adaptive response mechanisms. By leveraging historical interactions, user type, and product ownership data, the chatbot tailors its language, tone, and recommendations to individual users—enhancing engagement and support effectiveness. The final system is integrated with Statwolf’s service infrastructure through secure APIs, enabling access to real-time product data, service histories, and maintenance records. Continuous monitoring and retraining mechanisms allow the chatbot to improve over time based on user feedback and new data. Evaluation of system performance focuses on multiple dimensions: intent recognition accuracy, response relevance, user satisfaction, and system efficiency. Both automated metrics and human evaluations are used to validate the chatbot’s effectiveness in real-world scenarios. In summary, the adopted methodology delivers a robust, intelligent, and adaptive chatbot system. By combining advanced NLP, machine learning, and knowledge engineering techniques, the project provides an innovative solution that enhances Statwolf’s customer support and after-sales services through personalized, AI-driven interactions.
Machine Learning, Natural Language Processing