Zhenchao Jiang

Shenzhen Institutes of Advanced Technology

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Trends

Author Order

Document Type

Co-Authors

Degen Huang

Dalian University of Technology

Lishuang Li

Dalian University of Technology

Liuke Jin

Dalian University of Technology

Rui Guo

Jiangxi Transportation Research Institute

Liang Gu

Chongqing University of Technology

Qingshan Jiang

Yancheng Institute of Technology

Shun Guo

Shenzhen Institutes of Advanced Technology

Mingqing Huang

Fuzhou University

Jing Zhang

Renmin University of China

Yanchen Qiao

Peng Cheng Laboratory

Cooperative Institutions

Chinese Academy of Sciences

Dalian University of Technology

Shenzhen Institutes of Advanced Technology

University of Chinese Academy of Sciences

Zhejiang University of Technology

Chinese Academy of Medical Sciences & Peking Union Medical College

Xiamen University

Shenzhen Institute of Information Technology

Nanchang University

PRG S&Tech (South Korea)

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Field

Biomedical named entity recognition based on extended Recurrent Neural Networks

2021 IEEE International Conference on Bioinformatics and Biomedicine (BIBM) (2015)

Lishuang Li Liuke Jin Zhenchao Jiang Dingxin Song Degen Huang

Biomedical named entity recognition (bio-NER), which extracts important entities such as genes and proteins, has become one of the most fundamental tasks in biomedical knowledge acquisition. However, the performance of traditional NER systems is always limited to the construction of complex hand-designed features which are derived from various linguistic analyses and maybe only adapted to specified area. In this paper we mainly focus on building a simple and efficient system for bio-NER with the extended Recurrent Neural Network (RNN) which considers the predicted information from the prior node and external context information (topical information & clustering information). Extracting complex hand-designed features is skipped and replaced with word embeddings. The experiments conducted on the BioCreative II GM data set demonstrate RNN models outperform CRF model and deep neural networks (DNN); furthermore, the extended RNN model performs better than the original RNN model.

Named Entity Recognition

10.1109/bibm.2015.7359761

Cite

Citations (53)

Integrating Semantic Information into Multiple Kernels for Protein-Protein Interaction Extraction from Biomedical Literatures

PLoS ONE (2014)

Lishuang Li Panpan Zhang Tianfu Zheng Hongying Zhang Zhenchao Jiang

Protein-Protein Interaction (PPI) extraction is an important task in the biomedical information extraction. Presently, many machine learning methods for PPI extraction have achieved promising results. However, the performance is still not satisfactory. One reason is that the semantic resources were basically ignored. In this paper, we propose a multiple-kernel learning-based approach to extract PPIs, combining the feature-based kernel, tree kernel and semantic kernel. Particularly, we extend the shortest path-enclosed tree kernel (SPT) by a dynamic extended strategy to retrieve the richer syntactic information. Our semantic kernel calculates the protein-protein pair similarity and the context similarity based on two semantic resources: WordNet and Medical Subject Heading (MeSH). We evaluate our method with Support Vector Machine (SVM) and achieve an F-score of 69.40% and an AUC of 92.00%, which show that our method outperforms most of the state-of-the-art systems by integrating semantic information.

Tree kernel

Kernel (algebra)

Graph kernel

Tree (set theory)

Similarity (geometry)

10.1371/journal.pone.0091898

Cite

Citations (17)

A general protein-protein interaction extraction architecture based on word representation and feature selection

International Journal of Data Mining and Bioinformatics (2016)

Zhenchao Jiang shuang Li Degen Huang

Previous researches have shown that supervised Protein-Protein Interaction Extraction (PPIE) can get high accuracies with elaborately selected features and kernels. However, most features and kernels rest upon domain knowledge and natural language analysis, which makes the supervised model expensive, heavy and brittle. Moreover, commonly used representation techniques, such as one-hot encoding and Vector Space Model, fail to capture the semantic similarity between words. To reduce the manual labour and take advantage of semantic representation, we put forward a general instance representation architecture for PPIE, which integrates word representation, vector composition and feature selection. Our method obtains F-scores of 69.7, 78.8, 72.3, 72.0 and 83.7 on AIMed, BioInfer, HPRD50, IEPA and LLL respectively.

Representation

Similarity (geometry)

Feature vector

10.1504/ijdmb.2016.074878

Cite

Citations (17)

An approach to improve kernel-based Protein–Protein Interaction extraction by learning from large-scale network data

Methods (2015)

Lishuang Li Rui Guo Zhenchao Jiang Degen Huang

Representation

Kernel (algebra)

Code (set theory)

External Data Representation

10.1016/j.ymeth.2015.03.026

Cite

Citations (28)

Word Representation on Small Background Texts

Communications in computer and information science (2016)

Lishuang Li Zhenchao Jiang Yang Liu Degen Huang

Representation

Margin (machine learning)

10.1007/978-981-10-2993-6_12

Cite

Citations (2)

A Multi-channel Visualization Method for Malware Classification Based on Deep Learning

Yanchen Qiao Qingshan Jiang Zhenchao Jiang Liang Gu

The traditional malware classification method relies too much on expert extraction features, and the malware image visualization method contains fewer features. To deal with these problems, we propose a multi-channel visualization method for malware classification based on deep learning. Firstly, the malware binary file is divided into a 256×256-dimensional matrix according to the width of 256 bytes. Secondly, the Word2Vec algorithm is used to calculate the 256-dimensional vector of each byte in each binary file, and then the file is converted to a 256×256-dimensional matrix. Thirdly, we use the Word2Vec algorithm to calculate the 256-dimensional vector of each assembly instruction in each assembly file, and then the file is converted into a 256×256-dimensional matrix. Fourthly, for each malware sample, 3 matrixes are combined into an uncompressed multi-channel image. Finally, the LeNet5 is used for training classification model. The experimental results show that the average accuracy is 98.76%.

Word2vec

10.1109/trustcom/bigdatase.2019.00109

Cite

Citations (22)

Coreference resolution in biomedical texts

2021 IEEE International Conference on Bioinformatics and Biomedicine (BIBM) (2014)

Lishuang Li Liuke Jin Zhenchao Jiang Jing Zhang Degen Huang

Coreference resolution recently plays a more and more important role for many natural language processing tasks. In this paper, we propose two methods for the biomedical coreference resolution. One is the single machine learning method (SVM ranker-learning algorithm) which selects appropriate features for the pronoun and noun phrase coreference resolution respectively. The other one is the hybrid method which adopts the rule-based method or the machine learning method for relative pronouns, non-relative pronouns and noun phrases coreference resolution respectively. Experiments are carried out on Biomedical Natural Language Process Shared Task (BioNLP-ST) ¹ 2011 coreference resolution corpus. In the first method (the single machine learning method), the F-score is 49.36%, higher than that using the same method with the features in the Reconcile system by 10.06%. In the second method (the hybrid method), the F-score is 68.61%, higher than that of the currently best system by 1.21%.

Coreference

10.1109/bibm.2014.6999392

Cite

Citations (10)

Drug drug interaction extraction from literature using a skeleton long short term memory neural network

2021 IEEE International Conference on Bioinformatics and Biomedicine (BIBM) (2017)

Zhenchao Jiang Liang Gu Qingshan Jiang

Drug Drug Interactions (DDIs) can cause harmful effect. Two shared tasks, DDIExtraction 2011 and DDIExtraction 2013, have been held to promote the implementation and comparative assessment of natural language processing techniques in the field of the pharmacovigilance domain. However, few model can meanwhile achieve state-of-the-art performance on both tasks. A major reason is the lack of representation of DDI instance structure in common. Therefore, in this paper, we propose a novel method to make full use of the DDI structure based on deep learning, in which we grasp the skeleton structure of DDI instances by a skeleton long short term memory (skeleton-LSTM) network. The experimental results show that our method can achieve an F-score of 0.677 on DDIExtraction 2011 and an F-score of 0.714 on DDIExtraction 2013, both of which are state-of-the-art.

Skeleton (computer programming)

Representation

Drug-drug interaction

10.1109/bibm.2017.8217708

Cite

Citations (14)

Drug–drug interaction extraction from biomedical literature using support vector machine and long short term memory networks

Information Sciences (2017)

Degen Huang Zhenchao Jiang Li Zou Lishuang Li

Binary classification

Drug reaction

10.1016/j.ins.2017.06.021

Cite

Citations (60)

Phar-LSTM: a pharmacological representation-based LSTM network for drug–drug interaction extraction

PeerJ (2023)

Mingqing Huang Zhenchao Jiang Shun Guo

Pharmacological drug interactions are among the most common causes of medication errors. Many different methods have been proposed to extract drug-drug interactions from the literature to reduce medication errors over the last few years. However, the performance of these methods can be further improved. In this paper, we present a Pharmacological representation-based Long Short-Term Memory (LSTM) network named Phar-LSTM. In this method, a novel embedding strategy is proposed to extract pharmacological representations from the biomedical literature, and the information related to the target drug is considered. Then, an LSTM-based multi-task learning scheme is introduced to extract features from the different but related tasks according to their corresponding pharmacological representations. Finally, the extracted features are fed to the SoftMax classifier of the corresponding task. Experimental results on the DDIExtraction 2011 and DDIExtraction 2013 corpuses show that the performance of Phar-LSTM is competitive compared with other state-of-the-art methods. Our Python implementation and the corresponding data of Phar-LSTM are available by using the DOI 10.5281/zenodo.8249384.

Softmax function

Python

Representation

10.7717/peerj.16606

Cite

Citations (1)