14.20 Chatbot applications

#chatbot #NLP #AI #LLM

• Chatbots also known as conversational agents are systems designed to interact with humans via conversations
• chatbots fall into 2 categories:
  • Chatbots: these carry out extended conversation with the goal of mimicking the unstructured chats characters of informal human human communication
  • they are mostly used for fun
  • Goal based dialogue agents:
  • these are used to solve sepcific tasks like booking a flight or maintaining a shopping list

Chatbots are widely used across all domains these days:

1. customer service
2. healthcare
3. education
4. ecommerce
5. entertainment

Specific Applications and Usecases

• Personal assistant: these include things like SIRI, alexa, google assistant etc
• Simple and relatively short interactions like for playing music or setting alarms
• longer conversations are just for fun or practical applications
• Booking and travel reservations
• Clinical use for mental health coucelling including bots like wysa
• creative writing assistants like chatgpt
• FAQ like retrieval based chatbots
• structured workflows like the same
• fundamental recommendations
• personal asisstant
• creative tasks

Underlying Tech and Architecture:

• User Interface (UI): For text or voice input

• NLU (Natural Language Understanding): Analyzes user input to extract intent and entities identifying key terms and sepcific info like dates or names

• Dialogue Management (DM): Maintains the converstaion context and decides the next action

• Response Generation (RG): Generates the chatbot output

• Retrieval based chatbots: rely on a fixed set of responses from a database

  • they only identify user intent and retrieve the most appropriate prewritten response, they do not generate new content but match inputs to existing answers
  • these are typically used in closed domain scenarios like tasks for FAQ or structured workflows
  • Include methods like Rule Based matching and Ml models or decision trees for response selection
  • Embeddings like BERT can be used for similarity matching in retrieval based intent detection

• Generative Chatbots: Use advanced AI models to dynamically generate responses based on user input and context

  • powered by ML, NLP and DL including models like GPT or LSTM, they can handle open ended and complex questions