SmartFAQs.ai
Back to Learn
Intermediate

BERT

A transformer-based encoder architecture used in RAG for semantic search and relevance scoring; it utilizes bidirectional self-attention to generate context-aware embeddings, though it faces a computational trade-off between the high speed of Bi-Encoders for initial retrieval and the high precision of Cross-Encoders for re-ranking.

Definition

A transformer-based encoder architecture used in RAG for semantic search and relevance scoring; it utilizes bidirectional self-attention to generate context-aware embeddings, though it faces a computational trade-off between the high speed of Bi-Encoders for initial retrieval and the high precision of Cross-Encoders for re-ranking.

Disambiguation

Encoder-only architecture used for 'reading' and 'mapping' data, as opposed to GPT's Decoder-only architecture used for 'writing' text.

Visual Metaphor

"A 360-degree panoramic camera that captures the entire context of a sentence simultaneously to map its exact location in a multidimensional library."

Key Tools
Hugging Face TransformersSentence-Transformers (SBERT)PyTorchONNX Runtime
Related Connections

Conceptual Overview

A transformer-based encoder architecture used in RAG for semantic search and relevance scoring; it utilizes bidirectional self-attention to generate context-aware embeddings, though it faces a computational trade-off between the high speed of Bi-Encoders for initial retrieval and the high precision of Cross-Encoders for re-ranking.

Disambiguation

Encoder-only architecture used for 'reading' and 'mapping' data, as opposed to GPT's Decoder-only architecture used for 'writing' text.

Visual Analog

A 360-degree panoramic camera that captures the entire context of a sentence simultaneously to map its exact location in a multidimensional library.

Related Articles