Publications
See my publications below.
Cross‑type French Multiword Expression Identification with Pre‑trained Masked Language Models Van‑Tuan Bui and Agata Savary (2024) In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC‑COLING 2024), pages 4198–4204, Torino, Italia. ELRA and ICCL. [PDF]
We explore the fine-tuning of several state-of-the-art neural transformers for each MWE type. Our experiments demonstrate the advantages of the combined system over multi-task approaches or single-task models, addressing the challenges posed by diverse tagsets within the training data. Specifically, we evaluated the combined system on a French treebank named Sequoia, which features an annotation layer encompassing all syntactic types of French MWEs. With this combined approach, we improved the F1-score by approximately 3% on the Sequoia dataset.
Protocol Recognition in Virtual Avionics Network Based on Efficient and Lightweight Convolutional Neural Network M. Kerkech, V. ‑T. Bui, M. Africano, L. Martin and K. Srinivasarengan (2022) IEEE/ACM 26th International Symposium on Distributed Simulation and Real Time Applications (DS‑RT), Alès, France, 2022, pp. 9‑16, doi: 10.1109/DS‑RT55542.2022.9932068. [DOI]
In this work, we present AvioNet, a lightweight, computation-efficient neural network for virtual avionics network protocol recognition with accuracy and latency levels as required by aerospace systems. This method converts each packet into a common gray image, and then uses the depthwise separable convolution, pointwise group convolution and channel shuffle operations to automatically extract the appropriate spatial features. This reduces the computational complexity significantly while maintaining almost the same accuracy. This CNN-based classifier is verified on data that has non-avionic protocols mixed with avionic simulated protocols and is compared with the state-of-the-art methods. Experimental results show that the accuracy of the method exceeds 99.999% for avionics simulated dataset and outperforms other deep learning classifiers. Furthermore, the method provides low-latency guarantees that aerospace systems demand.
