My name is Xin Mao, supervised by Prof. Man Lan in School of Computer Science, East China Normal University(ECNU). My research interests include Knowledge Graph, Graph Representation Learning, Entity Alignment, and Information Extraction. Here are my CV and Publications. This is the last year of my Ph.D. career, thus I'm seeking an abroad post-doc/research fellow position. If there is a suitable position, please contact me by email, Thanks!
Majoring in computer science and technology and obtaining a bachelor's degree. My major in the first year was physics, thus I took 5 years to get my bachelor's degree.
I received the qualification of promotion and exemption with a decent GPA (5 / 120) and continue to learn at East China Normal University to pursue my doctorate.
Participated in the Alibaba air cooperation project (cross-lingual product attribute extraction).
Participated in the relation extraction and shop graph construction project.
n this paper, we argue that existing GNN-based EA methods inherit the inborn defects from their neural network lineage: weak scalability and poor interpretability. Inspired by recent studies, we reinvent the Label Propagation algorithm to effectively run on KGs and propose a non-neural EA framework -- LightEA. With a mere tenth of time consumption, LightEA achieves comparable results to state-of-the-art methods across all datasets and even surpasses them on many.
In this paper, we propose an effective and efficient EA decoding algorithm via third-order tensor isomorphism (DATTI). Extensive experiments on public datasets indicate that our decoding algorithm can deliver significant performance improvements even on the most advanced EA methods, while the extra required time is less than 3 seconds.
In this paper, we successfully transform the crosslingual EA problem into the assignment problem. Based on this finding, we propose a frustratingly Simple but Effective Unsupervised EA method (SEU) without neural networks. Experiments on widely used public datasets indicate that SEU outperforms all advanced competitors and has high efficiency, interpretability, and stability.
In this paper, we prove that the essence of GNNs-based EA methods is to solve a permutation matrix approximately and explain why negative samples are unnecessary from a new angle of view. Furthermore, we propose a novel EA method with three major modifications. The experimental results show that PSR not only surpasses the previous SOTA in performance but also has impressive scalability and robustness.
In this paper, we propose a novel KG encoder — Dual Attention Matching Network (Dual-AMN), which not only models both intra-graph and cross-graph information smartly, but also greatly reduces computational complexity. Furthermore, we propose the Normalized Hard Sample Mining Loss to smoothly select hard negative samples with reduced loss shift. The experimental results on widely used public datasets indicate that our method achieves both high accuracy and high efficiency.
In this paper, we raise the counter-intuitive phenomena in entity alignment, which are neglected by previous studies. By abstracting existing entity alignment methods into a unified framework, we successfully explain the questions and derive two key criteria for transformation operation in entity alignment. Inspired by these findings, we propose a novel GNNs-based method, Relational Reflection Entity Alignment (RREA).
This paper proposes a novel Meta Relation Aware Entity Alignment (MRAEA) to directly model cross-lingual entity embeddings by attending over the node's incoming and outgoing neighbors and its connected relations' meta semantics. Our experiments on all three benchmark entity alignment datasets show that our approach consistently outperforms the state-of-the-art methods.