Published in Arxiv, 2021
This is a book everything you need to know when you start to learning something about NLP. The book was compiled by Professors Tong Xiao and Jingbo Zhu, and I was responsible for the development of the LaTeX code.
Recommended citation:
Download Paper | Download Bibtex
Published in Workshop, 2024
This paper focuses on Huawei Translation Services Center’s (HW-TSC’s) submission to the sentence-level QE shared task.
Recommended citation: Shan W, Zhu M, Li Y, et al. HW-TSC 2024 Submission for the Quality Estimation Shared Task[C]//Proceedings of the Ninth Conference on Machine Translation. 2024: 535-540.
Download Paper
Published in Workshop, 2022
Our models are based on several advanced Transformer variants, e.g., Transformer-ODE, Universal Multiscale Transformer (UMST). The main workflow consists of data filtering, large-scale data augmentation (i.e., iterative back-translation, iterative knowledge distillation), and specific-domain fine-tuning.
Recommended citation: Zhang Y, Wang Z, Cao R, et al. The niutrans machine translation systems for wmt20[C]//Proceedings of the Fifth Conference on Machine Translation. 2020: 338-345.
Download Paper
Published in Workshop, 2022
WMT20 Rank first in Japanese↔English both side.
Recommended citation: Zhang Y, Wang Z, Cao R, et al. The niutrans machine translation systems for wmt20[C]//Proceedings of the Fifth Conference on Machine Translation. 2020: 338-345.
Download Paper
Published in EMNLP, 2025
We propose Prompt-aware Mixture (PaM) to enhance the Speech LLM that uses multiple audio encoders.
Recommended citation: **Shan W**, Li Y, Zhang Y, et al. Enhancing Speech Large Language Models with Prompt-Aware Mixture of Audio Encoders[J]. arXiv preprint arXiv:2502.15178, 2025.
Download Paper | Download Slides
Published in ACL 2025 Fingdings, 2025
We propose the unit language to overcome the two modeling challenges. The unit language can be considered a text-like representation format, constructed using n-gram language modeling.
Recommended citation: Zhang Y, Ma X, Kou K, et al. Leveraging Unit Language Guidance to Advance Speech Modeling in Textless Speech-to-Speech Translation[J]. arXiv preprint arXiv:2505.15333, 2025.
Download Paper
Published in ICASSP IEEE, 2024
We propose a novel generalized feature fusion framework grounded in conditional computation, mitigates feature conflicts and bolsters model robustness to multi-view input features.
Recommended citation: Shan W, Zhang Y, Han Y, et al. Optimizing Speech Multi-View Feature Fusion through Conditional Computation[C]//ICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2025: 1-5.
Download Paper | Download Slides
Published in ACL 2024 Fingdings, 2024
We introduce PartialFormer, a parameter-efficient Transformer architecture utilizing multiple smaller FFNs to reduce parameters and computation while maintaining essential hidden dimensions.
Recommended citation: Zhang Y, Ma X, Kou K, et al. Leveraging Unit Language Guidance to Advance Speech Modeling in Textless Speech-to-Speech Translation[J]. arXiv preprint arXiv:2505.15333, 2025.
Download Paper
Published in Arxiv, 2023
We explore the possibility of LLMs EE without additional output layers and joint optimization.
Recommended citation: Shan W, Meng L, Zheng T, et al. Early exit is a natural capability in transformer-based models: An empirical study on early exit without joint optimization[J]. arXiv preprint arXiv:2412.01455, 2024.
Download Paper
Published in Arxiv, 2023
Survey on research about early exit and mixture of expert.
Recommended citation: Shan W, et al. A Survey on Conditional Computation[J]. 2023.
Download Paper