<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Jnanitha Gotam on Medium]]></title>
        <description><![CDATA[Stories by Jnanitha Gotam on Medium]]></description>
        <link>https://medium.com/@gotamjnanitha?source=rss-43216877d677------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*mFWVUD1DgjGGg4ly</url>
            <title>Stories by Jnanitha Gotam on Medium</title>
            <link>https://medium.com/@gotamjnanitha?source=rss-43216877d677------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 13:41:08 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@gotamjnanitha/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Understanding RAG Retrieval: How AI Finds Relevant Information Before Answering]]></title>
            <link>https://medium.com/@gotamjnanitha/understanding-rag-retrieval-how-ai-finds-relevant-information-before-answering-34ee98fbaedf?source=rss-43216877d677------2</link>
            <guid isPermaLink="false">https://medium.com/p/34ee98fbaedf</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[generative-ai]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Jnanitha Gotam]]></dc:creator>
            <pubDate>Tue, 17 Nov 2026 12:07:25 GMT</pubDate>
            <atom:updated>2026-11-17T12:07:25.631Z</atom:updated>
            <content:encoded><![CDATA[<h4>1. Introduction</h4><p>Artificial Intelligence (AI) is becoming an important part of our daily lives. Large Language Models (LLMs) such as ChatGPT can understand questions and generate human-like answers. However, these models may sometimes provide incorrect information or fail to answer questions about specific documents.</p><p>Retrieval-Augmented Generation (RAG) is a technology that helps address this problem. It combines information retrieval with AI-generated responses.</p><p>Instead of depending only on the information learned during training, a RAG system searches relevant external information and uses it to generate an answer.</p><h3>2. What is Retrieval-Augmented Generation?</h3><p>Retrieval-Augmented Generation (RAG) is an AI technique that combines two main processes:</p><ul><li><strong>Retrieval:</strong> Finding relevant information from an external knowledge source.</li><li><strong>Generation:</strong> Using a Large Language Model (LLM) to generate an answer based on the retrieved information.</li></ul><p>For example, imagine uploading a university study material PDF and asking:</p><h4>“What is Artificial Intelligence?”</h4><p>A RAG system searches the uploaded PDF, finds the relevant information and sends it to an AI model. The model then generates an answer using that information.</p><h3>3. Why is RAG Needed?</h3><p>Traditional Large Language Models have some limitations:</p><ul><li>They may not know the latest information.</li><li>They may not have access to private documents.</li><li>They can generate incorrect or unsupported answers.</li><li>Updating their knowledge through retraining can be expensive.</li></ul><p>RAG helps by connecting language models to external knowledge sources. It allows users to retrieve information from documents, databases and other knowledge repositories without retraining the language model for every new document.</p><h3>4. How Does RAG Work?</h3><p>A typical RAG system follows these steps:</p><ol><li><strong>Document Collection:</strong> Collect information from PDFs, text files, websites or databases.</li><li><strong>Text Extraction:</strong> Extract readable text from the documents.</li><li><strong>Text Chunking:</strong> Divide large documents into smaller sections called chunks.</li><li><strong>Embedding Generation:</strong> Convert each chunk into a numerical representation called an embedding.</li><li><strong>Vector Database Storage:</strong> Store the embeddings in a vector database.</li><li><strong>User Query:</strong> The user enters a question.</li><li><strong>Similarity Search:</strong> Convert the question into an embedding and find the most relevant chunks.</li><li><strong>Answer Generation:</strong> Send the retrieved information and question to an LLM to generate a response.</li></ol><h3>5. Architecture of RAG</h3><p>A basic RAG architecture contains the following components:</p><ul><li><strong>Knowledge Source:</strong> Documents or other external information sources.</li><li><strong>Document Processor:</strong> Extracts and divides text into chunks.</li><li><strong>Embedding Model:</strong> Converts text into numerical vectors.</li><li><strong>Vector Database:</strong> Stores and searches vector embeddings.</li><li><strong>Retriever:</strong> Finds relevant information based on the user’s question.</li><li><strong>Large Language Model:</strong> Generates a natural-language response using the retrieved context.</li></ul><p>The retrieval process connects the user’s question to relevant information before the generation stage.</p><h3>6. Key Components of RAG</h3><h3>6.1 Large Language Model (LLM)</h3><p>An LLM is an AI model trained on large amounts of text. It understands language patterns and generates human-like responses.</p><h3>6.2 Embeddings</h3><p>Embeddings are numerical representations of text. They help computers compare the meaning of different pieces of text.</p><h3>6.3 Vector Database</h3><p>A vector database stores embeddings and supports similarity searches. FAISS is one example of a library used for efficient similarity search.</p><h3>6.4 Retriever</h3><p>The retriever searches the vector database and selects the information most relevant to the user’s question.</p><h3>6.5 Text Chunking</h3><p>Chunking divides long documents into smaller sections. This makes retrieval more focused and helps the system provide relevant context to the language model.</p><h3>7. Real-World Applications of RAG</h3><p>RAG can be used in different industries:</p><ul><li><strong>Education:</strong> Answering questions from textbooks, notes and study materials.</li><li><strong>Healthcare:</strong> Helping professionals retrieve relevant information from medical documents.</li><li><strong>Customer Support:</strong> Finding answers from product manuals and company knowledge bases.</li><li><strong>Business:</strong> Searching internal company reports and documents.</li><li><strong>Research:</strong> Finding relevant information from research papers and publications.</li></ul><h3>8. Advantages of RAG</h3><ul><li>Provides answers using relevant external information.</li><li>Can work with private or organization-specific documents.</li><li>Reduces dependence on information stored during model training.</li><li>Allows knowledge sources to be updated without retraining the LLM.</li><li>Can provide references to retrieved documents or passages.</li><li>Supports question answering across large collections of documents.</li></ul><h3>9. Limitations of RAG</h3><p>Although RAG is useful, it also has some limitations:</p><ul><li>Poor-quality documents can lead to poor answers.</li><li>Incorrect chunking can affect retrieval accuracy.</li><li>Embedding models may fail to capture some meanings.</li><li>Retrieval may return irrelevant information.</li><li>The system depends on the quality of the knowledge source.</li><li>Processing and storing large document collections can require additional resources.</li></ul><p>RAG does not completely eliminate incorrect AI answers. Its performance depends on retrieval quality, the source documents and the language model.</p><h3>10. Future Scope of RAG</h3><p>RAG technology continues to develop. Future improvements may include:</p><ul><li>More accurate retrieval techniques.</li><li>Better handling of large document collections.</li><li>Multimodal RAG that works with text, images, audio and other data.</li><li>Hybrid search combining keyword and semantic search.</li><li>Improved source citations and answer verification.</li><li>Integration with AI agents and automated workflows.</li></ul><p>These developments can make RAG systems more useful for education, research and business applications.</p><h3>11. Conclusion</h3><p>Retrieval-Augmented Generation is an important approach in modern AI because it connects language models with external knowledge sources.</p><p>It works by retrieving relevant information first and then generating an answer using that information. This makes it useful for document-based question answering and knowledge retrieval applications.</p><p>Understanding RAG provides a foundation for developing AI applications that can work with real-world documents and information.</p><h3>References</h3><p>Lewis et al., “Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks,” 2020.</p><p>LangChain Documentation — <a href="https://python.langchain.com/">https://python.langchain.co</a>m/</p><p>Hugging Face Documentation — <a href="https://huggingface.co/docs">https://huggingface.co/docs</a></p><p>FAISS Documentation — <a href="https://faiss.ai/">https://faiss.ai/</a></p><p>_________________________________________________________________</p><h3><em>✍️ Author</em></h3><p><strong><em>Jnanitha Gotam</em></strong></p><p><strong><em>Computing &amp; Data Science Student | Exploring AI, Machine Learning and Emerging Technologies.</em></strong></p><p><em>I am a B.Tech student at Sai University, passionate about Artificial Intelligence, Machine Learning, Data Science, and emerging technologies. I enjoy exploring new technologies, learning how AI works, and sharing my knowledge through simple and informative articles.</em></p><p><em>My goal is to continuously learn, build practical skills, and explore how technology can solve real-world problems.</em></p><p><strong><em>Connect with me:<br>GitHub: </em></strong><a href="https://github.com/gotamjnanitha-afk">https://github.com/gotamjnanitha-afk</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=34ee98fbaedf" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>