To protect patient privacy, medical models are frequently trained on synthetic datasets. Focusing on 13 clinical attributes commonly found in breast cancer records, we propose a synthetic data generation framework in which each generated sample is a data–report pair: the structured data capture the 13 attributes of interest, while the accompanying (French) text simulates a corresponding breast cancer pathology report. Data privacy and coverage are obtained by generating the structured data using a tabular data synthesiser. In this work, we focus on enhancing the diversity and quality of the generated texts through a combination of data-driven knowledge distillation, data filtering and data combination. Our findings suggest that the quality of the training data used for the student model outweighs its quantity; that combining data from multiple teachers helps improve performance; and that a compact model (Mistral 7B) can match the zero-shot Information Extraction performance of a significantly larger model (MedGemma 27B) when fine-tuned on high-quality synthetic data.

We describe the acquisition of a dialog corpus for French based on multi-task human-machine interactions in a serious game setting. We present a tool for data collection that is configurable for multiple games; describe the data collected using this tool and the annotation schema used to annotate it; and report on the results obtained when training a classifier on the annotated data to associate each player turn with a dialog move usable by a rule based dialog manager. The collected data consists of approximately 1250 dialogs, 10454 utterances and 168509 words and will be made freely available to academic and nonprofit research.

Previous work has shown that large scale subcategorisation lexicons could be extracted from parsed corpora with reasonably high precision. In this paper, we apply a standard extraction procedure to a 100 millions words parsed corpus of french and obtain rather poor results. We investigate different factors likely to improve performance such as in particular, the specific extraction procedure and the parser used; the size of the input corpus; and the type of frames learned. We try out different ways of interleaving the output of several parsers with the lexicon extraction process and show that none of them improves the results. Conversely, we show that increasing the size of the input corpus and modifying the extraction procedure to better differentiate prepositional arguments from prepositional modifiers improves performance. In conclusion, we suggest that a more sophisticated approach to parser combination and better probabilistic models of the various types of prepositional objects in French are likely ways to get better results.