Abstract:Retrieving and extracting knowledge from extensive research documents and large databases presents significant challenges for researchers, students, and professionals in today's information-rich era. Existing retrieval systems, which rely on general-purpose Large Language Models (LLMs), often fail to provide accurate responses to domain-specific inquiries. Additionally, the high cost of pretraining or fine-tuning LLMs for specific domains limits their widespread adoption. To address these limitations, we propose a novel methodology that combines the generative capabilities of LLMs with the fast and accurate retrieval capabilities of vector databases. This advanced retrieval system can efficiently handle both tabular and non-tabular data, understand natural language user queries, and retrieve relevant information without fine-tuning. The developed model, Generative Text Retrieval (GTR), is adaptable to both unstructured and structured data with minor refinement. GTR was evaluated on both manually annotated and public datasets, achieving over 90% accuracy and delivering truthful outputs in 87% of cases. Our model achieved state-of-the-art performance with a Rouge-L F1 score of 0.98 on the MSMARCO dataset. The refined model, Generative Tabular Text Retrieval (GTR-T), demonstrated its efficiency in large database querying, achieving an Execution Accuracy (EX) of 0.82 and an Exact-Set-Match (EM) accuracy of 0.60 on the Spider dataset, using an open-source LLM. These efforts leverage Generative AI and In-Context Learning to enhance human-text interaction and make advanced AI capabilities more accessible. By integrating robust retrieval systems with powerful LLMs, our approach aims to democratize access to sophisticated AI tools, improving the efficiency, accuracy, and scalability of AI-driven information retrieval and database querying.

What problem does this paper attempt to address?

The paper primarily addresses the existing issues in current information retrieval systems and proposes a novel solution. Specifically, the paper aims to solve the following two main problems: 1. **Limitations of Existing Retrieval Systems**: Current information retrieval systems often rely on general large-scale language models (LLMs), which may provide inaccurate or insufficient answers when handling domain-specific queries. Additionally, the high cost of pre-training or fine-tuning these models to adapt to specific domains limits their widespread application. 2. **Need for Efficient Storage and Retrieval of Natural Language Data**: With the rapid growth of information in academic literature and large databases, efficiently storing and retrieving natural language data has become particularly critical. However, current methods are often impractical for high-speed search operations on large-scale datasets. To address the above issues, the paper proposes a novel approach that combines generative AI and contextual learning. This approach constructs an advanced information retrieval system by integrating the capabilities of large-scale language models with the fast and accurate retrieval abilities of vector databases. The system can efficiently handle both structured and unstructured data, understand users' natural language queries, and retrieve relevant information without the need for fine-tuning. In this way, the researchers aim to overcome the limitations of existing systems, improve the efficiency, accuracy, and scalability of information retrieval, while reducing the high costs associated with model fine-tuning. The main contributions of the paper include: - Proposing a new method called "Generative Text Retrieval" (GTR), which can adapt to both structured and unstructured data and can be used with minimal adjustments. - Developing a variant called "Generative Tabular Text Retrieval" (GTR-T) to optimize the query process for structured databases, especially when handling large amounts of tabular data. - Demonstrating through experiments that the proposed models achieve excellent performance on multiple benchmark datasets, particularly achieving a 98% Rouge-L F1 score on the MSMARCO dataset, an execution accuracy (EX) of 0.82 on the Spider dataset, and an exact match (EM) accuracy of 0.60. In summary, this paper proposes an innovative information retrieval method aimed at improving the performance of retrieval systems, especially in handling domain-specific knowledge and large-scale datasets.

Enhancing Knowledge Retrieval with In-Context Learning and Semantic Search through Generative AI

WeKnow-RAG: An Adaptive Approach for Retrieval-Augmented Generation Integrating Web Search and Knowledge Graphs

Active Retrieval Augmented Generation

Empowering Large Language Models to Set up a Knowledge Retrieval Indexer via Self-Learning

Information retrieval from textual data: Harnessing large language models, retrieval augmented generation and prompt engineering

Retrieval-Augmented Generation for Large Language Models: A Survey

A Survey on Retrieval-Augmented Text Generation for Large Language Models

AI for the Generation and Testing of Ideas Towards an AI Supported Knowledge Development Environment

From Matching to Generation: A Survey on Generative Information Retrieval

Generative Information Retrieval Evaluation

Text Encoders Lack Knowledge: Leveraging Generative LLMs for Domain-Specific Semantic Textual Similarity

Towards Optimizing a Retrieval Augmented Generation using Large Language Model on Academic Data

Blended RAG: Improving RAG (Retriever-Augmented Generation) Accuracy with Semantic Search and Hybrid Query-Based Retrievers

DRAGIN: Dynamic Retrieval Augmented Generation based on the Information Needs of Large Language Models

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

Document Knowledge Graph to Enhance Question Answering with Retrieval Augmented Generation

Deploying Large Language Models With Retrieval Augmented Generation

Retrieval-Augmented Generation for AI-Generated Content: A Survey

Think-on-Graph 2.0: Deep and Faithful Large Language Model Reasoning with Knowledge-guided Retrieval Augmented Generation

ConTReGen: Context-driven Tree-structured Retrieval for Open-domain Long-form Text Generation