Kejian Liu

Mr. Kejian Liu

orcid

Research Focus & Metrics

Academic Career Length

15Years (2012-Present)

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

Representative Publications

View All Publications
31 January 2012 in Biological Trace Element Research
Verified
Journal Article
Closed access
PDF file
XLXiao-Li Liu
Xiao-Li Liu
CLChang-Cheng Li
Chang-Cheng Li
KLKejian Liu
Kejian Liu
CCCai-Yan Cui
Cai-Yan Cui
YZYu-Zeng Zhang
Yu-Zeng Zhang
YLYun Liu
Yun Liu

30 July 2023 in Electronics
Verified
Journal Article
Open access
PDF file
KLKejian Liu
Kejian Liu
YFYuanyuan Feng
Yuanyuan Feng
LZLiying Zhang
Liying Zhang
RWRongju Wang
Rongju Wang
WWWei Wang
Wei Wang
HLHailing Li
Hailing Li
While user-generated textual content on social platforms such as Weibo provides valuable insights into public opinion and social trends, the influence of personality on sentiment expression has been largely overlooked in previous studies, especially in Chinese short texts. To bridge this gap, we propose the P-BiLSTM-SA model, which integrates personalities into sentiment classification by combining BiLSTM and self-attention mechanisms. We grouped Weibo texts based on personalities and constructed a personality lexicon using the Big Five theory and clustering algorithms. Separate sentiment classifiers were trained for each personality group using BiLSTM and self-attention, and their predictions were combined by ensemble learning. The performance of the P-BiLSTM-SA model was evaluated on the NLPCC2013 dataset and showed significant accuracy improvements. In particular, it achieved 82.88% accuracy on the NLPCC2013 dataset, a 7.51% improvement over the baseline BiLSTM-SA model. The results highlight the effectiveness of incorporating personality factors into sentiment classification of short texts.

28 May 2019 in Toxicology Mechanisms and Methods
Verified
Journal Article
Closed access
HWHuiting Wang
Huiting Wang
XLXiang Li
Xiang Li
GZGe Zhao
Ge Zhao
LXLiangtao Xu
Liangtao Xu
SWSheng Wang
Sheng Wang
FXFuwei Xie
Fuwei Xie
The tobacco-specific nitrosamine 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK) is classified as a Group 1 human carcinogen. It is metabolically activated by P450 enzymes to intermediate methylate and pyridyloxobutylate DNA, resulting in the formation of DNA adduct that is critical for the carcinogenicity of NNK. To directly and objectively examine the DNA adduct formation profiles without the complexity of factors in vivo, in the present study, five kinds of methyl DNA adducts were first identified in the incubation model of NNK established with human lung epithelial cells (BEAS-2B). The level of methyl DNA adducts and metabolites of NNK were quantitatively analyzed, respectively. With the increase of exposure time and dose, the level of methyl DNA adducts and metabolites increased. Furthermore, with the changes of the activity of P450 enzymes, which is the main enzyme regulating the α-hydroxylation of NNK, we found the levels of both methyl adducts and metabolites formed via α-hydroxylation in experimental groups showed the same trend compared with those in control group, while the metabolites formed via other pathways changed in the opposite trend. The result proves that the methyl adducts induced by NNK generate via α-hydroxylation pathway in BEAS-2B cells.

1 January 2017 in International Journal of Computer Theory and Engineering
Verified
Journal Article
Closed access
LFLing Feng
Ling Feng
KLKejian Liu
Kejian Liu
FTFuxi Tang
Fuxi Tang
QMQingrui Meng
Qingrui Meng

8 June 2020 in Annalen der Physik
Verified
Journal Article
Closed access
PDF file
ZZZhe Zhang
Zhe Zhang
WJWeimin Jiang
Weimin Jiang
KLKejian Liu
Kejian Liu
MLMingrui Liu
Mingrui Liu
JMJianchao Meng
Jianchao Meng
JNJiacai Nie
Jiacai Nie

28 June 2023 in Applied Sciences
Verified
Journal Article
Open access
PDF file
KLKejian Liu
Kejian Liu
WWWei Wang
Wei Wang
RWRongju Wang
Rongju Wang
XCXuran Cui
Xuran Cui
LZLiying Zhang
Liying Zhang
XLXianyong Li
Xianyong Li
Modelling both long- and short-term user interests from historical data is crucial for generating accurate recommendations. However, unifying these metrics across multiple application domains can be challenging, and existing approaches often rely on complex, intertwined models which can be difficult to interpret. To address this issue, we propose a lightweight, plug-and-play interest enhancement module that fuses interest vectors from two independent models. After analyzing the dataset, we identify deviations in the recommendation performance of long- and short-term interest models. To compensate for these differences, we use feature enhancement and loss correction during training. In the fusion process, we explicitly split long-term interest features with longer duration into multiple local features. We then use a shared attention mechanism to fuse multiple local features with short-term interest features to obtain interaction features. To correct for bias between models, we introduce a comparison learning task that monitors the similarity between local features, short-term features, and interaction features. This adaptively reduces the distance between similar features. Our proposed module combines and compares multiple independent long-term and short-term interest models on multiple domain datasets. As a result, it not only accelerates the convergence of the models but also achieves outstanding performance in challenging recommendation scenarios.