Conversational agents based on Large Language Models (LLMs) have recently emerged as powerful tools for human-computer interaction. Nevertheless, their black-box nature implies challenges in predictability and a lack of personalization, both of which can be addressed by controlled generation. This work proposes an end-to-end method to obtain modular and explainable control over LLM outputs through ontological definitions of aspects related to the conversation. Key aspects are modeled and used as constraints; we then further fine-tune the LLM to generate content accordingly. To validate our approach, we explore two tasks that tackle two key conversational aspects: the English proficiency level and the polarity profile of the content. Using a hybrid fine-tuning procedure on seven state-of-the-art, open-weight conversational LLMs, we show that our method consistently outperforms pre-trained baselines, even on smaller models. Beyond quantitative gains, the framework remains model-agnostic, lightweight and interpretable, enabling reusable control strategies that can be extended to new domains and interaction goals. This approach enhances alignment with strategy instructions and demonstrates the effectiveness of ontology-driven control in conversational systems.

Generative AI has simplified information access by enabling natural language-driven interactions between users and automated systems. In particular, Question Answering (QA) has emerged as a key application of AI, facilitating efficient access to complex information through dialogue systems and virtual assistants. The Large Language Models (LLMs) combined with Knowledge Graphs (KGs) have further enhanced QA systems, allowing them to not only correctly interpret natural language but also retrieve precise answers from structured data sources such as Wikidata and DBpedia. However, enabling LLMs to generate machine-readable SPARQL queries from natural language questions (NLQs) remains challenging, particularly for complex questions. In this study, we present experiments in fine-tuning LLMs for the task of NLQ-to-SPARQL transformation. We rely on benchmark datasets for training and testing the fine-tuned models, generating queries directly from questions written in English (without further processing of the input or output). By conducting an analytical study, we examine the effectiveness of each model, as well as the limitations associated with using fine-tuned LLMs to generate SPARQL.