Hongzhi Zhang

Mr. Hongzhi Zhang

orcid

Research Focus & Metrics

Academic Career Length

10Years (2017-Present)

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

Representative Publications

View All Publications
27 November 2018 in IEEE Access
Verified
Journal Article
Open access
HZHongzhi Zhang
Hongzhi Zhang
GXGuandong Xu
Guandong Xu
XLXiao Liang
Xiao Liang
GXGuangluan Xu
Guangluan Xu
FLFeng Li
Feng Li
THTinglei Huang
Tinglei Huang
Relation detection is a challenging task because of the high variance of relation expression in real-world questions, and it plays a crucial role in Knowledge Base Question Answering (KBQA). Traditional relation detection models based on deep learning follow an encoding-comparing paradigm, where the question and the candidate relation are represented as vectors to compare their semantic similarity. Max- or average- pooling operation, which is used to compress the sequence of words into fixed-dimensional vectors, becomes the bottleneck of information flow. In this paper, we propose an Attention-Based Wordlevel Interaction Model (ABWIM) to alleviate the information loss issue caused by aggregating the sequence into a fixed-dimensional vector before the comparison. Firstly, attention mechanism is adopted to learn the soft alignments between words from the question and the relation. Then fine-grained comparisons are performed on the aligned words. Finally, the comparison results are merged with a simple recurrent layer to estimate the semantic similarity. Besides, a dynamic sample selection strategy is proposed to accelerate the training procedure without decreasing the performance. Experimental results of relation detection on both SimpleQuestions and WebQuestions datasets show that ABWIM achieves state-of-the-art accuracy, demonstrating its effectiveness.

1 November 2019 in Knowledge-Based Systems
Verified
Journal Article
Closed access
HZHongzhi Zhang
Hongzhi Zhang
GXGuangluan Xu
Guangluan Xu
XLXiao Liang
Xiao Liang
WZWeili Zhang
Weili Zhang
XSXian Sun
Xian Sun
THTinglei Huang
Tinglei Huang

14 September 2023
Verified
Conference Paper
Open access
SWSirui Wang
Sirui Wang
PLPeiguang Li
Peiguang Li
YXYunsen Xian
Yunsen Xian
HZHongzhi Zhang
Hongzhi Zhang
The sequential recommendation task aims to predict the item that user is interested in according to his/her historical action sequence. However, inevitable random action, i.e. user randomly accesses an item among multiple candidates or clicks several items at random order, cause the sequence fails to provide stable and high-quality signals. To alleviate the issue, we propose the StatisTics-Driven Pre-traing framework (called STDP briefly). The main idea of the work lies in the exploration of utilizing the statistics information along with the pre-training paradigm to stabilize the optimization of recommendation model. Specifically, we derive two types of statistical information: item co-occurrence across sequence and attribute frequency within the sequence. And we design the following pre-training tasks: 1) The co-occurred items prediction task, which encourages the model to distribute its attention on multiple suitable targets instead of just focusing on the next item that may be unstable. 2) We generate a paired sequence by replacing items with their co-occurred items and enforce its representation close with the original one, thus enhancing the model’s robustness to the random noise. 3) To reduce the impact of random on user’s long-term preferences, we encourage the model to capture sequence-level frequent attributes. The significant improvement over six datasets demonstrates the effectiveness and superiority of the proposal, and further analysis verified the generalization of the STDP framework on other models.

5 June 2018 in Entropy
Verified
Journal Article
Open access
PDF file
HZHongzhi Zhang
Hongzhi Zhang
XLXiao Liang
Xiao Liang
GXGuangluan Xu
Guangluan Xu
KFKun Fu
Kun Fu
FLFeng Li
Feng Li
THTinglei Huang
Tinglei Huang
Automatic question answering (QA), which can greatly facilitate the access to information, is an important task in artificial intelligence. Recent years have witnessed the development of QA methods based on deep learning. However, a great amount of data is needed to train deep neural networks, and it is laborious to annotate training data for factoid QA of new domains or languages. In this paper, a distantly supervised method is proposed to automatically generate QA pairs. Additional efforts are paid to let the generated questions reflect the query interests and expression styles of users by exploring the community QA. Specifically, the generated questions are selected according to the estimated probabilities they are asked. Diverse paraphrases of questions are mined from community QA data, considering that the model trained on monotonous synthetic questions is very sensitive to variants of question expressions. Experimental results show that the model solely trained on generated data via the distant supervision and mined paraphrases could answer real-world questions with the accuracy of 49.34%. When limited annotated training data is available, significant improvements could be achieved by incorporating the generated data. An improvement of 1.35 absolute points is still observed on WebQA, a dataset with large-scale annotated training samples.

26 May 2026 in ACM Transactions on Intelligent Systems and Technology
Verified
Journal Article
Open access
ZZZhuo Zhang
Zhuo Zhang
JZJingyuan Zhang
Jingyuan Zhang
JHJingtao Huang
Jingtao Huang
HWHui Wang
Hui Wang
YYYue Yu
Yue Yu
ZXZenglin Xu
Zenglin Xu
Instruction tuning has been identified as a crucial technique for optimizing large language models (LLMs) to generate human-aligned responses. Nonetheless, gathering diversified and superior-quality instruction data for such tuning presents notable obstacles, especially in privacy-sensitive domains. Federated instruction tuning ( FedIT ) has emerged as a promising solution by consolidating collaborative training across multiple data owners, resulting in a privacy-enhancing learning model. Existing FedIT studies assume that clients have sufficient training data, however, in reality, many clients only have few-shot samples, leading to either overfitting in federated LLM or degraded performance. At the same time, this federated few-shot environment also increases the risk of training data extraction attacks, as the LLM may well memorize the limited training data. To address these issues, this paper proposes a novel federated algorithm, PPFedIT , designed to enhance privacy protection and model performance of federated few-shot learning. PPFedIT comprises three vital steps on the client side: (1) synthetic data generation, which utilizes the strong generation capacity of LLMs to generate synthetic data, aiming to diversify and enrich local data; (2) parameter isolation training, which respectively updates the parameters of a shared global LLM on the synthetic data and the parameters of local LLMs on the local data, consequently mitigating the noise impact of the synthetic data; (3) local aggregation then sharing mechanism, which mixes the parameters of the global LLM and those of a local LLM first, before uploading them to a server for aggregation. This effectively mitigates data extraction attacks. Extensive experiments on three open-source datasets demonstrate PPFedIT significantly enhances model performance (averaging 8.4%) and reduces the risk of data extraction attacks (approximately 20%) in practical and challenging federated few-shot scenarios.

22 May 2017 in Information
Verified
Journal Article
Open access
PDF file
HZHongzhi Zhang
Hongzhi Zhang
WZWeili Zhang
Weili Zhang
THTinglei Huang
Tinglei Huang
XLXiao Liang
Xiao Liang
KFKun Fu
Kun Fu
The intensive construction of domain-specific knowledge bases (DSKB) has posed an urgent demand for researches about domain-specific entity detection and linking (DSEDL). Joint models are usually adopted in DSEDL tasks, but data imbalance and high computational complexity exist in these models. Besides, traditional feature representation methods are insufficient for domain-specific tasks, due to problems such as lack of labeled data, link sparseness in DSKBs, and so on. In this paper, a two-stage joint (TSJ) model is proposed to solve the data imbalance problem by discriminatively processing entity mentions with different degrees of ambiguity. In addition, three novel methods are put forward to generate effective features by incorporating an unlabeled corpus. One crucial feature involving entity detection is the mention type, extracted by a long short-term memory (LSTM) model trained on automatically annotated data. The other two types of features mainly involve entity linking, including the inner-document topical coherence, which is measured based on entity co-occurring relationships in the corpus, and the cross-document entity coherence evaluated using similar documents. An overall 74.26% F1 value is obtained on a dataset of real-world movie comments, demonstrating the effectiveness of the proposed approach and indicating its potentiality to be used in real-world domain-specific applications.