Shunyu Yao

Mr. Shunyu Yao

orcid

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 January 2022
Verified
Conference Paper
Open access
PDF file
YSYisi Sang
Yisi Sang
XMXiangyang Mou
Xiangyang Mou
MYMo Yu
Mo Yu
SYShunyu Yao
Shunyu Yao
JLJing Li
Jing Li
JSJeffrey Stanton
Jeffrey Stanton
Yisi Sang, Xiangyang Mou, Mo Yu, Shunyu Yao, Jing Li, Jeffrey Stanton. Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2022.

5 July 2024
Verified
Conference Paper
Open access
SYShunyu Yao
Shunyu Yao
QKQingqing Ke
Qingqing Ke
QWQiwei Wang
Qiwei Wang
KLKangtong Li
Kangtong Li
JHJie Hu
Jie Hu
The emergence of large language models has brought about revolutionary changes in the field of natural language processing and has shown extraordinary potential in general tasks and various specific domain tasks, especially in the legal field. However, there are still many factors that constrain the application of large language models in the legal field, with the main problems being the lack of domain knowledge and the ability to apply knowledge to solve problems. Therefore, we propose Lawyer GPT, a legal large model that incorporates domain knowledge by using an external knowledge retrieval module to combine an external knowledge base and possesses legal reasoning capabilities. We have collected a large amount of legal domain data and combined it with general domain data, using GPT-4 Turbo to build a high-quality legal dataset. To make Lawyer GPT's legal reasoning capabilities more reliable, we have performed supervised fine-tuning on this dataset, providing it with a good ability to apply domain knowledge to solve problems and enabling it to independently handle various legal professional issues. In addition, we have constructed a legal knowledge base and used retrieval enhancement techniques to provide Lawyer GPT with tools to retrieve external domain knowledge, thereby improving the factual and rationality of the generated content. Experimental results show that Lawyer GPT has demonstrated good performance in both subjective and objective legal domain tests, showing a strong ability to handle legal issues.

1 September 2021 in Knowledge-Based Systems
Verified
Journal Article
Closed access
SYShunyu Yao
Shunyu Yao
KSKai Shuang
Kai Shuang
RLRui Li
Rui Li
SSSen Su
Sen Su


29 November 2024 in Electronics
Verified
Journal Article
Open access
PDF file
SYShunyu Yao
Shunyu Yao
DLDan Liu
Dan Liu
YSYingming Shi
Yingming Shi
ZZZhiyuan Zhang
Zhiyuan Zhang
JHJie Hu
Jie Hu
With the rapid development of e-commerce, malicious accounts have become a threat, especially in the business field. Therefore, how to efficiently detect malicious accounts has become one an important issue requiring resolution. Present research on malicious detection mainly uses the basic statistics of the original data to build models, and ignores malicious activities in the business field. To address the context-independent nature of business activities and their lack of social structure, this study constructs an online model for detecting malicious accounts in the business field based on a stacking ensemble strategy with BiGRU-Conv1D-Capsule, XGBoost, and LightGBM as individual learners and AdaBoost as a meta-learner. Experimental results show that the stacking ensemble model constructed in this study has better predictive power than the typical shallow learning and baseline deep learning models. In addition to the basic feature, the behavior sequence of users is also applied in the proposed model. Overall, our experiments based on a real dataset from a financial platform show that the TPR, AUC, and F1 of the stacking ensemble model can reach 0.8258, 0.9860, and 0.8922 respectively, which outperforms all baseline models.

5 July 2024
Verified
Conference Paper
Open access
SYShunyu Yao
Shunyu Yao
QKQingqing Ke
Qingqing Ke
KLKangtong Li
Kangtong Li
QWQiwei Wang
Qiwei Wang
JHJie Hu
Jie Hu
With the continuous development of natural language processing (NLP) technology, large models have become essential tools for handling natural language tasks. In the news domain, large models can be used for automating news generation, thereby enhancing the productivity and quality of news production. As a result, we introduce a new large model—News GPT—which utilizes an external knowledge retrieval module to inject real information, providing authentic and reliable news generation services. By combining Chinese and English news data with general domain data, we have constructed a high-quality, multi-domain news dataset consisting of 1.4B tokens. In the pre-training phase, we expand the Chinese vocabulary for the Llama2-70B model, and in the fine-tuning phase, we design expert prompts to help the model understand downstream tasks better. We have tested News GPT in various aspects, and the experimental results show that News GPT, as an intelligent news assistant, can effectively complete tasks using retrieval tools in different application scenarios. It possesses excellent natural language understanding, knowledge, and logical reasoning abilities, and effectively controls the hallucinations in the generation process. News GPT demonstrates high accuracy and reliability in news generation and can provide substantial support for the news industry.