School of Engineering Department of Computer Science and Engineering 135 Commonsense Reasoning with Knowledge Graphs Supervisor: SONG Yangqiu / CSE Student: KWONG Hyun Fung / COMP Course: UROP 1100, Fall In this research, we try to apply commonsense reasoning in solving high-school level physics word questions. We aim to build a knowledge graph-based model and apply commonsense reasoning to tackle physics word problems. These problems will require the model to understand the context of the questions, as well as apply suitable formulas. To build a suitable question bank as the dataset for the model, we have design multiple physics simulators to generate a physics word questions dataset with various types of questions under different conditions. This will require the model to apply different formulas base on the question. The research is still in progress, and we are going to build a problem formulator to transform the physics word questions into queries, which could be put into the model. Our aim is to build a model that could apply commonsense reasoning and use suitable formulas to solve these questions. Graph Machine Learning for Logical Reasoning Supervisor: SONG Yangqiu / CSE Student: CHEN Rulin / MATH-GM Course: UROP 1100, Fall This report summaries the work done during the 2024 fall semester. The UROP project mainly focuses on using logical reasoning methods for the Complex Query Answering (CQA) problem. We use the knowledge graph to model the entities and the relationships between them, and based on the knowledge graph, we try to use some graph neural network approaches to answer complex queries raised by the knowledge graph. We also analyze the datasets (both the knowledge graph and the queries) used for training and hope to get some insights on improving the model performance by proposing a new dataset sampling method and then training or fine-tuning on the new dataset to overcome the possible bias problem in existing dataset. Graph Machine Learning for Logical Reasoning Supervisor: SONG Yangqiu / CSE Student: TAN Rickson Caleb Yap / DSCT Course: UROP 1100, Fall Complex Logical Query Answering (CLQA) is an emerging area of interest in graph machine learning that addresses the complex task of multi-hop logical reasoning over large, potentially incomplete graphs. This paper will present two ideas: a way to transform graph databases with the concept of Neural Graph Databases (NGDB) and another is a current state-of-the-art model for CLQA on standard Knowledge Graphs by Query Computation Tree Optimization. NGDBs enhance standard Graph Databases and consist of a Neural Graph Storage and a Neural Graph Engine. Leveraging NGDBs, QTO achieves optimal embeddings by passing them through a tree-like computation graph known as the Query Computation Tree, representing the current leading model for CLQA.
RkJQdWJsaXNoZXIy NDk5Njg=