Dongsheng Guo

Dongsheng Guo

Auto-generated profile, based on publicly available information

Dongsheng Guo

Inspur Academy of Science and Technology, China

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

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

Representative Publications

View All Publications
13 November 2020
Verified
Book Chapter
Closed access
PDF file
DGDongsheng Guo
Dongsheng Guo
HLHongzhi Liu
Hongzhi Liu
HZHaoru Zhao
Haoru Zhao
YCYunhao Cheng
Yunhao Cheng
QSQingwei Song
Qingwei Song
BZBing Zheng
Bing Zheng

22 December 2023
Verified
Conference Paper
Open access
ZWZizhong Wei
Zizhong Wei
DGDongsheng Guo
Dongsheng Guo
DHDengrong Huang
Dengrong Huang
QZQilai Zhang
Qilai Zhang
SZSijia Zhang
Sijia Zhang
RLRui Li
Rui Li
Large language models (LLMs) have achieved impressive success in generating fluent and coherent texts in natural language. However, the presence of inaccurate or low-quality data can unintentionally lead to the retention of incorrect knowledge, resulting in hallucinations that hinder progress in content generation. In this paper, we propose a comprehensive framework aimed at detecting and mitigating these hallucinations. Our approach uses Named Entity Recognition (NER) and Entity Relationship (ER) models to identify hallucination entities and sentences during the detection phase. Furthermore, by incorporating prompt engineering, we effectively correct these hallucination sentences using LLM in the mitigation phase. Tests on real articles confirm the effectiveness of our approach in rectifying LLM-associated hallucinations without adding new ones, thereby enhancing their reliability and credibility.

22 December 2023
Verified
Conference Paper
Open access
KJKai Jiang
Kai Jiang
QZQilai Zhang
Qilai Zhang
DGDongsheng Guo
Dongsheng Guo
DHDengrong Huang
Dengrong Huang
SZSijia Zhang
Sijia Zhang
RLRui Li
Rui Li
News generation, as a subset of long-form text generation task, is an important challenge in natural language processing. The creation of news articles necessitates a foundation in factual events, demanding the authenticity of content. The current large language models (LLMs) are capable of generating fluent text. However, these models often result in the generation of text that includes unfounded or inaccurate information, a phenomenon known as hallucinations. To ease this issue, we introduce a novel method. This approach leverages structured prompt-template inputs to guide the generation process, coupled with a post-checking module to ensure content authenticity. Furthermore, we propose two novel metrics, namely, topic consistency and core consistency, to evaluate the quality of the generated news. We have conducted extensive experiments using real news datasets, and our findings indicate that the proposed framework demonstrably mitigates the incidence of hallucinations in algorithmically generated news content, concurrently enhancing the overall quality of the produced news articles.

24 May 2024
Verified
Conference Paper
Open access
GWGuangxin Wang
Guangxin Wang
XLXue Li
Xue Li
DGDongshen Guo
Dongshen Guo
QDQiang Duan
Qiang Duan
QCQibin Chen
Qibin Chen
ZWZizhong Wei
Zizhong Wei
Identifying and preventing unsafe behaviors is an important task to protect people’s lives and property. Due to the complexity of behavioral types and the fact that many behavioral actions have small amplitude, rendering ineffective spatial and temporal variation information, zero-shot image-based detection of unsafe behaviors is a highly challenging task. The multimodal large language model shows excellent performance in the zero-shot task once it appears. However, it is often impossible to accurately locate and analyze each individual’s behavior in complex scenes solely relying on its ability. Therefore, this paper proposes an innovative zero-shot method to detect unsafe behaviors. We built a two-stage detection model. First, we focus on the target area of people through the target detection algorithm, and then analyze the behavior through the multimodal large language model combined with the construction strategy of positive and negative description prompts. This method shows excellent performance on the dataset we built. It can accurately identify unsafe behaviors in the real environment, which is very important to protect public safety and property.

22 October 2022 in International Journal of Radiation Oncology - Biology - Physics (IJROBP)
Verified
Journal Article
Closed access
PDF file
XYX. Ye
X. Ye
DGD. Guo
D. Guo
JLJ. Liu
J. Liu
JGJ. Ge
J. Ge
HYH. Yu
H. Yu
SKS. Kong
S. Kong

24 May 2024
Verified
Conference Paper
Open access
DGDongsheng Guo
Dongsheng Guo
ZWZizhong Wei
Zizhong Wei
RLRui Li
Rui Li
XLXue Li
Xue Li
KJKai Jiang
Kai Jiang
QDQiang Duan
Qiang Duan
Video matting is a core but arduous task in computer vision and digital image processing. Our research specifically targets the improvement of human matting performance in the hand region, a notoriously difficult aspect due to factors, e.g., motion blur. To address this, we propose an innovative video matting approach that uniquely detects and estimates the landmark structure of the hand, focusing on the fingers, then employs morphological dilation techniques around these landmarks to bolster the α matte values, thereby enhancing the extraction quality of the hand region. Through experiments conducted using a custom-built evaluation dataset consisting of videos with hand movements, we demonstrate that our method significantly improves hand segmentation by effectively enhancing the α matte of the hands.