Search papers, labs, and topics across Lattice.
This paper introduces TextQ-German, a comprehensive dataset suite designed for the human-centered evaluation of Natural Language Generation (NLG) in German, focusing on Quality of Experience (QoE). By conducting crowdsourced studies, the authors collect human quality ratings and identify key perceptual dimensions relevant to automatic text summarization and machine translation tasks. The study reveals that hybrid QoE prediction models, which integrate linguistic features with transformer-based approaches, significantly outperform traditional methods, thereby establishing a valuable resource for enhancing NLG systems to align more closely with human quality perceptions.
Hybrid models that blend linguistic features with transformer architectures can significantly enhance the evaluation of German NLG, outperforming traditional baselines.
The rapid advancement of Natural Language Generation (NLG) has made the reliable evaluation of generated text increasingly critical, as these systems, such as large language models (LLMs), are now widely deployed in real-world applications. However, traditional automatic metrics fail to capture the multifaceted nature of perceived quality. In this paper, we introduce TextQ-German, a novel dataset suite for human-centered evaluation of German NLG from a Quality of Experience (QoE) perspective, covering automatic text summarization and machine translation. Through crowdsourcing studies with German speakers, we collect human quality ratings and identify relevant perceptual quality dimensions for each task. We develop automatic QoE prediction models, including transformer-based, linguistic feature-based, and hybrid approaches. Hybrid models outperform pure transformer baselines in almost all experimental settings, while linguistic features alone can approach the performance of fine-tuned language models. The dataset is extended with LLM-generated outputs annotated with overall QoE scores. Final validation on held-out sets indicates generalization to unseen data. Our work contributes a publicly accessible resource for NLG evaluation and baselines for automatic QoE prediction, providing a foundation for developing NLG systems that better align with human quality perception.
