Stroke is among the most common causes of death and disability worldwide. More than 12 million people suffer from a stroke every year, 4 million of whom develop an aphasia. Aphasia is a language disorder that affects up to 0.6\% of the global population. Broca's aphasia is a form of aphasia that affects the ability to articulate speech, but not the ability to understand it. It has a negative impact on the mental health and quality of life of people who suffer from it. Patients may suffer from communication difficulties in daily life. A tool to help patients with aphasia in their daily communication could be useful. In this scope, our contribution in this paper is twofold. First, we propose a method to convert correct French sentences into sentences containing errors that could have been produced by aphasic people. For that, we use lexical errors produced by ChatGPT. We use this method to create a synthetic parallel corpus of French sentences and their Broca's aphasia equivalents. Second, we propose a transformer-based translation model to map aphasic sentences back to their correct form. Our model is trained on the synthetic parallel corpus and achieves a BLEU score of 79.61.