Definition:
- Semantic search is a search technique that improves search accuracy by better capturing the contextual meaning of search terms, rather than relying solely on keyword matching.
Key Features:
- Contextual Understanding: It employs natural language processing (NLP) to analyze the context and meaning of the query, matching it to relevant information based on semantic similarities rather than just keyword matches.
- Embedding-Based Retrieval: Queries and documents are transformed into dense vector embeddings, often using models like BERT or GPT. These embeddings capture semantic similarities.
- Similarity Matching: Search results are retrieved based on the similarity of embeddings, using techniques like cosine similarity.
Use Cases:
- Search Engines: Improved search result relevance.
- Recommendation Systems: Better content recommendations based on user queries.
- Customer Support: Enhanced FAQ and document retrieval systems.
Definition:
- RAG is a method that combines retrieval-based methods with generative models to improve the generation of responses or content by grounding them in retrieved information.
Key Features:
- Two-Stage Process:
- Retrieval: Relevant documents or pieces of information are retrieved from a large corpus using techniques similar to those in semantic search.
- Generation: A generative model, like GPT, uses the retrieved information to produce a coherent and contextually accurate response.
- Integration of Retrieval and Generation: Strives that generated content is accurate and grounded in real data.
Use Cases:
- Question Answering: Generating precise answers based on retrieved documents.
- Content Generation: Producing articles, reports, or summaries that are grounded in factual data.
-
Focus:
- Semantic Search: Focuses on improving the accuracy of information retrieval by understanding query context.
- RAG: Focuses on generating accurate and contextually relevant content by combining retrieval and generation.
-
Methodology:
- Semantic Search: Uses embedding-based similarity matching for retrieval.
- RAG: Combines retrieval with generative models to produce responses.
-
Output:
- Semantic Search: Returns a list of relevant documents or pieces of information.
- RAG: Produces a coherent and contextually grounded response or piece of content.
Definition:
- A conversational agent, also known as a chatbot, is designed to interact with users through natural language conversations, providing information, and answering questions by leveraging the information with which the model was trained on.
Key Features:
- Two-Stage Process:
- Natural Language Understanding: Converts the natural language query into a vector representation that captures the contextual meaning based on patterns learned during pretraining. This allows the model to match the query to relevant knowledge stored in its parameters.
- Natural Language Generation: Generates responses by transforming the vector representation back into natural language, producing coherent and contextually appropriate answers based on learned associations.
- Contextual Conversations: Maintains and utilizes context within a conversation to ensure responses are coherent, relevant, and aligned with the flow of the dialogue, enhancing the user experience.
Use Cases:
- Customer Support: Answering customer queries, providing information.
- Personal Assistance: Answering general knowledge questions.
-
Core Function:
- RAG: Focuses on combining retrieval and generation to provide grounded, contextually relevant textual responses.
- Conversational Agent: Focuses on engaging in natural language conversations to provide information, answer questions, and track conversation, session, or user history context.
-
Methodology:
- RAG: Retrieves relevant documents or information based on the query and uses generative models to produce a response that integrates the retrieved information.
- Conversational Agent: Uses natural language understanding to process and interpret user queries and natural language generation to create responses, maintaining context throughout the conversation.
-
Output:
- RAG: Produces detailed, contextually grounded, and informative responses by integrating retrieved information with generative model capabilities.
- Conversational Agent: Generates interactive, conversational responses that are coherent and contextually appropriate, often incorporating session or user history to maintain continuity in dialogue.
-
Use Cases:
- Conversational Agent: Customer support, personal assistance, interactive services.
- RAG: Question answering, content creation, enhancing chatbot responses with grounded information.
Definition:
- A copilot is an AI assistant designed to assist users in specific domains, typically focusing on enhancing productivity and supporting complex workflows through suggestions, automations, and contextual assistance.
Key Features:
- Contextual Assistance: Provides suggestions and guidance based on the context of the user’s work.
- Domain-Specific Knowledge: Tailored to assist in specific domains such as coding, document editing, or project management.
- Interactive Support: Offers interactive help, including real-time suggestions and automations.
Capabilities:
- Task Assistance: E.g. offering code completions, document editing suggestions.
- Workflow Enhancement: E.g. streamlining project management tasks, automating repetitive tasks.
- Contextual Guidance: E.g. providing real-time tips and recommendations based on current activities.
Use Cases:
- Software Development: Assisting with code writing, debugging, and optimization.
- Content Creation: Helping with writing, editing, and formatting documents.
- Project Management: Offering task prioritization, deadline reminders, and workflow automation.
-
Core Function:
- Conversational Agent: Focuses on engaging in natural language conversations to provide information, answer questions, and track conversation, session, or user history context.
- Copilot: Focuses on enhancing productivity and supporting complex workflows through suggestions, automations, and contextual assistance tailored to specific domains.
-
Methodology:
- Conversational Agent: Uses natural language understanding to process and interpret user queries and natural language generation to create responses, maintaining context throughout the conversation.
- Copilot: Utilizes contextual analysis and domain-specific knowledge to provide real-time suggestions, automations, and guidance based on the user’s current tasks and activities.
-
Integration:
- Conversational Agent: Primarily leverages its parametric memory, using the information stored during its training to generate responses without external data retrieval.
- Copilot: Can use an adapted form of Retrieval-Augmented Generation (RAG) to provide user, activity, and domain-specific context into the underlying language model, integrating real-time information and external data sources to enhance its suggestions and automations.
-
Output:
- Conversational Agent: Generates interactive, conversational responses that are coherent and contextually appropriate, often incorporating session or user history to maintain continuity in dialogue.
- Copilot: Produces actionable suggestions, automations, and contextual guidance to enhance productivity and streamline workflows, specific to the user’s current domain and tasks.
-
Use Cases:
- Conversational Agent: Customer support, personal assistance, interactive services.
- Copilot: Software development, content creation, project management.
Definition:
- A virtual assistant capable of performing a range of actions beyond information retrieval, including executing commands, interacting with other systems, and automating tasks.
Key Features:
- Task Execution: Can perform specific tasks such as running database queries, creating calendar entries, booking flights, and more.
- Integration with External Systems: Interacts with various APIs and systems to perform actions, including semantic search retrieval for performing RAG.
- Intent Understanding: Uses NLP to infer user intent and execute appropriate actions.
Capabilities:
- Database Queries: Can execute SQL or other queries to retrieve and manipulate data.
- Calendar Management: Can create, modify, and delete calendar entries.
- Booking Services: Can book flights, hotels, and other services.
- Automation: Automates routine tasks and workflows based on user commands.
Use Cases:
- Personal Assistance: Managing schedules, sending reminders, booking appointments.
- Business Automation: Automating business processes, running reports, managing databases.
- Customer Service: Providing interactive support, executing service requests, and processing transactions.
-
Core Function:
- Copilot: Focuses on enhancing productivity and supporting complex workflows through domain-specific suggestions, automations, and contextual assistance.
- Agentic Virtual Assistant: Focuses on understanding user commands and performing a variety of tasks and actions beyond just providing information.
-
Capabilities:
- Copilot: Provides task assistance, workflow enhancement, and contextual guidance tailored to specific domains like software development, content creation, and project management.
- Agentic Virtual Assistant: Capable of executing a wide range of actions such as running queries, managing calendars, booking services, and more.
-
Integration:
- Copilot: Integrates with domain-specific tools and systems to provide relevant suggestions and automations based on the user's current tasks.
- Agentic Virtual Assistant: Integrates with various external systems and APIs to perform diverse tasks, including leveraging semantic search for retrieval-augmented generation (RAG).
-
Output:
- Copilot: Produces actionable suggestions, automations, and contextual guidance to enhance productivity and streamline workflows specific to the user's domain.
- Agentic Virtual Assistant: Executes specific tasks and actions, potentially providing a variety of outputs, including textual responses, system updates, and service bookings.
-
Use Cases:
- Copilot: Assisting with code writing, debugging, and optimization in software development; helping with writing, editing, and formatting documents in content creation; streamlining project management tasks.
- Agentic Virtual Assistant: Managing schedules, sending reminders, booking appointments for personal assistance; automating business processes, running reports, managing databases for business automation; providing interactive support, executing service requests, and processing transactions for customer service.
Recommendation: Don't focus on the specific tools, frameworks, and libraries used in the below resources, but on what they do, how they do it, and why they do it. Focus on first principles, and on architectural and workflow patterns.
- Large Language Models with Semantic Search
- Full Course
- Minimal Recommended Lectures:
- Optional:
- Generating Answers: This generation step makes it RAG.
- Building and Evaluating Advanced RAG
Full Course
- Minimal Recommended Lectures:
- Optional:
- RAG Triad of Metrics: RELEVANT FOR QA
- Sentence-Window Retrieval RELEVANT FOR DATA ENGINEERING (INGESTION & PREPROCESSING)
- Auto-Merging Retrieval RELEVANT FOR DATA ENGINEERING (INGESTION & PREPROCESSING)
- Building Agentic RAG with Llamaindex
- Full Course
- Minimal Recommended Lectures:
- Preprocessing Unstructured Data for LLM Applications
- Full Course
- Minimal Recommended Lectures:
- Optional:
- The 5 Levels Of Text Splitting For Retrieval
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent