Automatic text simplification for French: model fine-tuning for simplicity assessment and simpler text generation

Abstract Automatic text simplification models face the challenge of generating outputs that, while being indeed simpler, still retain some complexity. This stems from the inherently relative nature of simplification, wherein a given text is transformed into a relatively simpler version, which does not necessarily equate to simple. We thus aim to propose a finer-grained method to assess sentence complexity in French. Our solution comprises three models, in which two address absolute and relative sentence complexity assessment, while the third focuses on measuring simplicity gain. By employing this triad of models, we aim to offer a comprehensive approach to qualify and quantify sentence simplicity. Our approach utilizes FlauBERT, fine-tuned for classification and regression tasks. Based on our three-dimensional complexity analysis, we provide the WiViCo dataset, comprising 46,525 aligned complex-simpler pairs, which is further leveraged to fine-tune different FLAN-T5-based language models for simplified text generation. In this context, we perform different evaluation tasks that contrast human evaluations with BLEU and SARI metrics for the generated simplifications, the models’ computational efficiency and environmental impact.