NLP and AI in Chatbots Flashcards

What is 'Named Entity Recognition' (NER) used for in chatbots?

Answer: Automatically identifying and classifying key information like people, places, and dates in text

NER allows chatbots to extract structured facts from free-form user input to use in downstream logic.

How does a transformer-based model like BERT improve chatbot NLP?

Answer: By understanding context bidirectionally, capturing word meaning based on surrounding text

BERT reads text in both directions simultaneously, enabling much deeper contextual understanding than earlier sequential models.

What is 'slot filling' in task-oriented chatbots?

Answer: Collecting all required pieces of information (slots) from the user to complete a specific task

Slot filling guides the conversation to gather every required parameter before executing an action, like booking a flight.

What is 'dialogue state tracking' in a conversational AI system?

Answer: Maintaining a representation of what has been established in the conversation so far

Dialogue state tracking keeps a running record of confirmed intents, entities, and context to enable coherent multi-turn conversations.

What distinguishes a 'generative' chatbot from a 'retrieval-based' chatbot?

Answer: Generative bots produce new text responses using language models; retrieval bots select from predefined answers

Generative models create responses dynamically, offering flexibility, while retrieval-based models pick the best pre-written response from a fixed set.

What is 'zero-shot learning' in the context of chatbot NLP?

Answer: The model's ability to handle new intents it was never explicitly trained on by generalizing from context

Zero-shot learning leverages large pre-trained models to recognize new categories without requiring labeled examples for every intent.