✳flâneur — a map of the web's best reading

Aman's AI Journal • NLP • Retrieval Augmented Generation

aman.ai · 42,355 words · saved by 1 readers

A potential solution could be to combine the strengths of both databases: indexing parsed entity relationships with vector representations in a graph database for more flexible information retrieval. It remains to be seen if such a hybrid model exists. The image below (source), gives a visual overview of the three different steps of RAG: Ingestion, Retrieval, and Synthesis/Response Generation. In the sections below, we will go over these key areas. Building a RAG system involves determining the ideal chunk sizes for the documents that the retriever component will process. The ideal chunk size depends on several factors: Data Characteristics: The nature of your data is crucial. For text documents, consider the average length of paragraphs or sections. If the documents are well-structured with distinct sections, these natural divisions might serve as a good basis for chunking. Retriever Constraints: The retriever model you choose (like BM25, TF-IDF, or a neural retriever like DPR) might

Overview Motivation The Retrieval Augmented Generation (RAG) Pipeline Advantages of RAG RAG vs. Fine-tuning Ensemble of RAG Choosing a Vector DB using a Feature Matrix Building a RAG pipeline Ingestion Chunking Figuring out the ideal chunk size Retriever Ensembling and Re-ranking Embeddings Naive Chunking vs. Late Chunking vs. Late Interaction (ColBERT and ColPali) Overview Naive/Vanilla Chunking What is Naive/Vanilla Chunking? Example Advantages and Limitations Late Chunking What is Late Chunking? How Late Chunking Works Example Advantages and Trade-offs Late Interaction What is Late Interact

Explore this link on the map →

related reading