Summary of G-sap: Graph-based Structure-aware Prompt Learning Over Heterogeneous Knowledge For Commonsense Reasoning, by Ruiting Dai et al.
G-SAP: Graph-based Structure-Aware Prompt Learning over Heterogeneous Knowledge for Commonsense Reasoning
by Ruiting Dai, Yuqiao Tan, Lisi Mo, Shuang Liang, Guohao Huo, Jiayi Luo, Yao Cheng
First submitted to arxiv on: 9 May 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed G-SAP (Graph-based Structure-Aware Prompt Learning) model is designed to improve commonsense question answering by balancing heterogeneous knowledge and enhancing cross-modal interaction between Language Models (LM) and Knowledge Graphs (KG). The model combines ConceptNet, Wikipedia, and Cambridge Dictionary to create an evidence graph, which is then used to generate prompts driven by graph entities and relations. A structure-aware frozen PLM incorporates structured and textual information from the evidence graph, while a heterogeneous message-passing reasoning module facilitates deep interaction between LM and graph-based networks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The G-SAP model aims to overcome limitations of fully fine-tuned pre-trained LMs in commonsense reasoning by incorporating knowledge from multiple sources. The proposed approach outperforms existing models, including the SoTA LM+GNNs model, on three benchmark datasets with a notable 6.12% improvement on the OpenbookQA dataset. |
Keywords
* Artificial intelligence * Prompt * Question answering
