ACL is the top international conference in the field of natural language processing which is held once annually in the world. ACL is rated as Class A conference by China Computer Federation. According to Google Scholar Metrics in 2022 , the H5 index of ACL is 169 and H5 median is 304.

Li Jiang, the first author of two articles accepted by ACL 2023, is the PhD candidate supervised by the research team of Prof. Gao Guanglai. Two articles mainly focus on two key problems in intelligent question-answering: knowledge graph completion and intention understanding, which realizes the new breakthrough in natural language processing by IMU. The research is supported by the National Natural Science Foundation of China and Inner Mongolia Autonomous Region Program for Key Research and Development and Research Achievement Transformation. And the National & Local Joint Engineering Research Center of Intelligent Information Processing Technology for Mongolian provides the support of software and hardware for the research.

1. TeAST: Temporal Knowledge Graph Embedding via Archimedean Spiral Timeline

Authors: Li Jiang, Su Xiangdong, Gao Guanglai

Institution authors are with: Inner Mongolia University

A brief introduction to the article: In the temporal-knowledge-graph-based system, the temporal knowledge graph embedding(TKGE) model is often applied in inference-missing to infer and make decisions. The existing TKGE method integrates time information into entities to lead to the potential evolution of entity information and limit the link prediction performance of TKG. Meanwhile, the current TKGE model lacks the capacity to model important relations and provide interpretability and hinders its effectiveness and potential application. To eliminate the limitations, this article proposes a new TKGE model which through Archimedean spiral timeline and temporal knowledge graph embedding representation maps the relation onto the corresponding Archimedean spiral timeline and change the quadruplet completion into third order tenor completion. To be more specific, Archimedean spiral timeline ensures that the same relations are on the same timeline and all the relations change as time goes on. Furthermore, this article presents a new temporal spiral regularization function to make timeline ordered. In addition, the research provides mathematical proof to prove the coding capacity of TeAST for various relation models.

2. How Well Apply Simple MLP to Incomplete Utterance Rewriting?

Authors: Li Jiang, Su Xiangdong, Ma Xinlan, Gao Guanglai

Institution  authors are with: Inner Mongolia University

A brief introduction to the article: The purpose of Incomplete Utterance Writing(IUR) is to restore the incomplete sentences with adequate context information for understanding. This article introduces a simple and effective method of IUR-MIUR. Being different from the previous researches, this research, first of all, use mono-layer MLP architecture to tap the potential semantic information of the union sentences. Then the research uses eigen-matrices to predict the category of token to restore incomplete utterances. The well-designed network and simple architecture make the newly proposed method better than the existing methods in quality and inference speed.