HighlightCurated by Aramai Editorial

Construction of Knowledge Graphs: Current State and Challenges

A research paper on the current state and challenges of constructing knowledge graphs.

The authors discuss the main graph models for knowledge graphs, introduce requirements for future construction pipelines, and evaluate the state-of-the-art. They identify areas in need of further research and improvement. The paper provides an overview of necessary steps to build high-quality knowledge graphs, including metadata management and quality assurance.

Based on: Construction of Knowledge Graphs: Current State and Challenges · Information

HighlightCurated by Aramai Editorial

BioLORD-2023: semantic textual representations fusing large language models and clinical knowledge graph insights

A study on using large language models to complement biomedical knowledge graphs for training semantic models.

The authors propose a new approach for obtaining high-fidelity representations of biomedical concepts and sentences. They introduce BioLORD-2023, a state-of-the-art model for semantic textual similarity and biomedical concept representation designed for the clinical domain. The study demonstrates consistent improvements over previous state-of-the-art models across various datasets.

Based on: BioLORD-2023: semantic textual representations fusing large language models and clinical knowledge graph insights · Journal of the American Medical Informatics Association

HighlightCurated by Aramai Editorial

Domain-Specific Retrieval-Augmented Generation Using Vector Stores, Knowledge Graphs, and Tensor Factorization

A paper introducing SMART-SLIC, a domain-specific LLM framework integrating RAG with KG and vector store.

The authors present SMART-SLIC, a framework that combines retrieval-augmented generation (RAG) with knowledge graphs (KG) and vector stores to improve question answering accuracy in specific domains. The framework is designed to be generalizable and adaptable to various specialized domains. It aims to mitigate hallucinations, reduce fine-tuning needs, and attribute information sources.

Based on: Domain-Specific Retrieval-Augmented Generation Using Vector Stores, Knowledge Graphs, and Tensor Factorization

HighlightCurated by Aramai Editorial

Enhancing Retrieval-Augmented Generation Models with Knowledge Graphs: Innovative Practices Through a Dual-Pathway Approach

A research paper proposing a dual-pathway approach to enhance retrieval-augmented generation models using knowledge graphs.

The authors present a novel method for improving retrieval-augmented generation models by incorporating knowledge graphs. This approach involves a dual-pathway framework that combines the strengths of both retrieval and generation components. The proposed method is evaluated on several benchmarks, demonstrating its effectiveness in enhancing model performance.

Based on: Enhancing Retrieval-Augmented Generation Models with Knowledge Graphs: Innovative Practices Through a Dual-Pathway Approach · Lecture notes in computer science

HighlightCurated by Aramai Editorial

Knowledge Graph Reasoning and Security Assurance Decision-Making Based on Online Retrieval Augment Generation

A paper proposing a framework for enhancing security assurance using Knowledge Graph reasoning and online Retrieval Augmented Generation.

The authors present a novel approach to risk assessment and mitigation in critical infrastructure, leveraging Knowledge Graphs and large language models. The framework integrates a dynamically updated Knowledge Graph with LLMs to facilitate real-time risk evaluation and proactive strategies. Simulated experiments demonstrate the efficacy of this framework in improving risk identification and response.

Based on: Knowledge Graph Reasoning and Security Assurance Decision-Making Based on Online Retrieval Augment Generation

HighlightCurated by Aramai Editorial

From human experts to machines: An LLM supported approach to ontology and knowledge graph construction

Paper exploring the semi-automatic construction of Knowledge Graphs using Large Language Models.

The authors propose a pipeline for constructing Knowledge Graphs with minimal human involvement, leveraging open-source Large Language Models. They demonstrate their method on a deep learning methodology dataset. The paper evaluates the generated content and suggests that LLMs can reduce human effort in KG construction.

Based on: From human experts to machines: An LLM supported approach to ontology and knowledge graph construction · arXiv (Cornell University)

HighlightCurated by Aramai Editorial

Knowledge Graph Prompting for Multi-Document Question Answering

A method for formulating context in prompting large language models for multi-document question answering.

The authors propose a Knowledge Graph Prompting (KGP) method to improve multi-document question answering. KGP consists of graph construction and traversal modules, which create a knowledge graph over multiple documents and navigate across nodes to gather supporting passages. The method aims to enhance prompt design and retrieval augmented generation for large language models.

Based on: Knowledge Graph Prompting for Multi-Document Question Answering · Proceedings of the AAAI Conference on Artificial Intelligence

HighlightCurated by Aramai Editorial

HybridRAG: Integrating Knowledge Graphs and Vector Retrieval Augmented Generation for Efficient Information Extraction

A novel approach to enhance question-answer systems for information extraction from financial documents.

The paper introduces HybridRAG, a combination of Knowledge Graph-based RAG techniques and VectorRAG techniques. It aims to improve information extraction from financial documents by retrieving context from both vector databases and knowledge graphs. Experiments show that HybridRAG outperforms traditional VectorRAG and GraphRAG in terms of retrieval accuracy and answer generation.

Based on: HybridRAG: Integrating Knowledge Graphs and Vector Retrieval Augmented Generation for Efficient Information Extraction

HighlightCurated by Aramai Editorial

TrumorGPT: Query Optimization and Semantic Reasoning over Networks for Automated Fact-Checking

A generative AI solution for automated fact-checking that merges machine learning with natural language processing techniques.

The paper introduces TrumorGPT, a novel framework for automated fact-checking. It leverages a large language model with few-shot learning and retrieval-augmented generation to access updated knowledge graphs. This approach aims to combat misinformation by providing accurate and reliable information promptly.

Based on: TrumorGPT: Query Optimization and Semantic Reasoning over Networks for Automated Fact-Checking

HighlightCurated by Aramai Editorial

A Retrieval-Augmented Generation Approach for Data-Driven Energy Infrastructure Digital Twins

Paper presenting a data-driven energy digital-twin framework and architecture.

The paper proposes a novel data-driven and knowledge-based energy digital-twin framework, integrating machine learning with a knowledge graph to support a retrieval-augmented generation approach. This enhances a conversational virtual assistant for user decision support in asset management and maintenance. A prototype framework was implemented using commercial-off-the-shelf tools and tested on a case study.

Based on: A Retrieval-Augmented Generation Approach for Data-Driven Energy Infrastructure Digital Twins · Smart Cities

HighlightCurated by Aramai Editorial

REANO: Optimising Retrieval-Augmented Reader Models through Knowledge Graph Generation

A paper proposing a knowledge graph generation module to enhance retrieval-augmented reader models.

The authors propose REANO, a system that generates knowledge graphs from passages and uses them to improve the performance of retrieval-augmented reader models. This is done by adding a knowledge graph generator and an answer predictor to the model. Experimental results show improvements in exact match scores on five open domain question answering datasets.

Based on: REANO: Optimising Retrieval-Augmented Reader Models through Knowledge Graph Generation

HighlightCurated by Aramai Editorial

Enhancing Distractor Generation for Multiple-Choice Questions with Retrieval Augmented Pretraining and Knowledge Graph Integration

A research paper on improving distractor generation in multiple-choice questions using retrieval augmented pretraining and knowledge graph integration.

This paper proposes a method to enhance distractor generation in multiple-choice questions by combining retrieval augmented pretraining and knowledge graph integration. The approach aims to improve the quality of distractors, which are essential for assessing students' understanding. Experimental results demonstrate the effectiveness of the proposed method in generating high-quality distractors.

Based on: Enhancing Distractor Generation for Multiple-Choice Questions with Retrieval Augmented Pretraining and Knowledge Graph Integration