Natural Language Processing - Langchain

Flashcards for Natural Language Processing Lecture 8 on Langchain

Langchain

A framework designed to assist in building applications powered by large language models (LLMs).

Retrieval Augmented Generation (RAG)

A method that provides hints or supporting text within a prompt to guide the model's response, ensuring it relies on provided information.

Indexing Pipeline

A process that prepares documents for semantic search, involving text extraction, splitting, embedding, and storing vectors.

Text Extraction

The initial step in the indexing pipeline, converting various input formats (docx, pdf, images, voice) into text.

Text Splitting

Dividing extracted text into smaller segments using various options like size, overlapping, and structure for better processing.

Embedding

Converting text segments into numerical vectors, capturing their semantic meaning for similarity comparisons.

Vector Store

A database for storing and efficiently retrieving vector embeddings, enabling semantic similarity searches.

RAG QA Pipeline

A process involving question embedding, semantic similarity retrieval, prompt template filling, and running prompts through LLMs to generate answers.

Semantic Similarity Retrieval

Finding relevant documents or text segments in a vector store based on the meaning of a question.

Langflow

A tool for the visualization of Langchain architectures.

LangGraph

A tool for the visualization of Langchain architectures.