With the rapidly advancing field of Artificial Intelligence (AI), chatbots have emerged as powerful tools for enhancing customer engagement, automating tasks, and refining user experiences. As AI and chatbot use becomes more widespread, there will be lots of terminology that people may not be familiar with. Therefore, a chatbot and AI glossary of terminology will help serve as a valuable resource in helping to provide clarity and consistency of communication when discussing these topics.
Here is a list of some of the common glossary associated with AI and chatbots:
Adaptive Behaviour: ability of artificial intelligence (AI) systems to adjust or modify their actions, performance, and models based on new information.
AI (Artificial Intelligence) Chatbots: computer programs designed to engage user in human-like manner to accomplish an objective. They are also known as conversational agents or virtual agents.
Algorithm: sequence of instructions that enables an artificial intelligence system (e.g. chatbot) to perform a specific task or achieve a specific objective.
Algorithm Discrimination: refers to situations where ML and AI system exhibit biased or unfair outcomes towards certain groups or individuals because of biased training data or inadequate development processes.
Annotation: process of labelling data with additional information to help machine learning (ML) algorithms understand and learn.
Application Programming Interface (API): a set of rules, protocols and tools that allow different software applications to communicate with each other. It defines methods and data formats that applications can use to request and exchange information.
Artificial Intelligence (AI): describes various computing techniques and algorithms designed to simulate or mimic human intelligence and thought processes.
Bard: conversational Generative AI chatbot developed by Google. Based on LaMDA family of Large Language Models (LLMs). Rebranded to Gemini – a more powerful and multi-modal AI.
Black Box AI: AI systems where internal processes and decision-making are not easily interpretable or understandable.
Bot-Builders: a user-friendly application designed to help non-programmers create, design, and deploy chatbots. These applications allow users to define behaviour, responses, and functionality of the chatbot without the need for extensive programming knowledge.
Bot Framework: set of tools, libraries, and resources that developers use to build and deploy chatbots or conversational agents.
Chatbot Persona: refers to the identity or personality that a chatbot adopts to engage with the users in a human-like manner. It involves defining the chatbot’s tone and style to create a conversational experience that aligns with the brand identity of an organization.
Chatbot Widget: an interface that allows the user to interact with a chatbot. Widgets can be integrated into website or applications and can be customized to provide multiple features for the user.
Context Awareness: information retained by a chatbot from previous interactions in a conversation, enabling it to understand the user’s current input in context.
Controllability: ability to understand, regulate and manage AI system’s decision-making to ensure accuracy, safety and minimize potential for unwanted behaviours.
Conversational AI: a field of AI that focuses on developing systems that conduct human-like conversations.
Data Stewardship: refers to management, security, privacy, oversight, and responsible handling of data within an organization.
Deep Learning: a branch of machine learning (ML) and AI that focuses on using artificial neural networks to model and solve complex problems. It involves training on multiple layers to learn and make decisions from vast amounts of data.
Dialog Builders: provides designers and developers the flexibility to change and update chatbot interactions through a user-friendly application.
Entity: key pieces of information or data that the chatbot identifies and uses to fulfill a user’s request.
Explainability: refers to being able to understand and explain the actions and decisions made by AI systems.
Fallback Intent: a predefined response or action that a chatbot uses when it is unable to determine the user’s intent.
Frequently Asked Questions (FAQs): collections of most common questions and answers. It is a primary data source that chatbots use to pull information to complete knowledge tasks.
Gemini: a family of AI models developed by Google. It can understand and generate text, as well as understand information from images, audio, videos, and code.
Generative AI: category of AI systems that can generate new content or data. These systems can create images, text, music, or other types of content.
Generative Pre-trained Transformers (GPT): a family of neural network models developed by OpenAI. They are trained on massive amounts of data (millions to billions of parameters) and can be used to generate text, translate languages, and write different kinds of content.
Grounding: process of aligning the knowledge and understanding of an AI system with the real-world data it interacts with.
Guardrails: refers to rules or guidelines put in place to ensure responsible use of AI systems. These guardrails are implemented to prevent the AI from generating harmful outputs.
Hallucination: refers to machine learning (ML) model generations that are incorrect, or unrelated to the information present in training data.
Human Fallback: process of chatbot handing off a case to a human agent when it cannot resolve a user request.
Human in the Loop: a design or process where human intervention or oversight is integrated into an AI system to provide guidance, validation, or correction as needed.
Intent: refers to the purpose or goals that a user expresses during their input. Intent identification is important in building chatbots, as it allows the AI system to determine what the user is seeking.
Intent-Entity Framework: interaction is structured around identifying user intents and extracting specific entities to generate appropriate responses.
Intent Recognition: ability of a chatbot to accurately identify user intents based on their input.
Intentless Chatbots: type of conversational AI chatbot that operates through a Large Language Model (LLM) without predefined user intents. It has the ability to engage in more open-ended conversations and respond dynamically based on context and user input.
LaMDA: Language Model for Dialogue Applications, it is a family of Large Language Models developed by Google.
LLaMa: is a family of open-source Large Language Models built by Meta AI.
Large Language Models (LLMs): refers to complex models that are trained on massive amounts of data (millions to billions of parameters) and are designed to understand and generate human-like text.
Live Chat Integration: the ability of a chatbot to seamlessly hand over a conversation to a human agent for more complex or sensitive requests.
Machine Learning (ML): a branch of AI that involves the development of algorithms that enable computers to learn, improve from and make predictions based on data. This is one of the key components that help distinguish conversational AI-based chatbots from basic rule-based chatbots.
Multimodal Large Language Models: trained on large amounts of data – textual and non-textual – to generate responses in text, images, audio, or video.
Multi-turn Conversations: conversation involving multiple exchanges between a user and a chatbot to accomplish a task or answer a request.
Natural Language Generation (NLG): process of creating grammatically correct sentences through algorithms.
Natural Language Processing (NLP): involves the development of algorithms, models and systems that enables computers to process human language – either through speech or text. This is key component that distinguishes conversational AI-based chatbots from basic rule-based chatbots.
Natural Language Understanding (NLU): a component of NLP that enables chatbots to read and interpret human language to identify the user intent.
Neural Networks: inspired by the structure and functioning of the human brain, these are a class of algorithms designed to recognize patterns and make intelligent decisions based on data. These networks can learn from data and generalize patterns.
Omni-channel: refers to a business strategy that provides customers with consistent and integrated experience across multiple channels (e.g. online, mobile, etc.).
Open-Source AI: AI projects and frameworks whose source code is freely accessed, used, modified, and distributed to the public. Open-source AI models encourage collaboration, transparency, and community involvement in its development.
Platform Integration: the integration of a chatbot into various websites, applications, or messaging platforms for user interaction.
Pre-training: training a model on a large dataset before fine-tuning it to specific tasks.
Prompt Engineering: refers to strategic and deliberate crafting of input prompts to help guide LLMs to a desirable outcome.
Reinforcement Learning: type of ML in which the model learns to make decisions by receiving feedback.
Response generation: process by which a chatbot creates and delivers a reply based on user input.
Responsible AI: approach of creating, implementing, and deploying AI systems that foster trust, ethics, transparency, fairness, accountability, and focus on the well-being of individuals and society.
Retrieval Augmented Generation (RAG): process of enhancing accuracy and reliability of an LLM by grounding the model on external sources of knowledge to supplement its training data.
Rule-Based Chatbot: a computer program that works using a predetermined dialogue flow. Unlike AI chatbots, it can only operate within a rigid structure.
Sentiment Analysis: process of determining emotional tone expressed in user’s input to enable the chatbot to respond appropriately.
Scripted Responses: responses that are predetermined based on anticipated user inputs.
Supervised Learning: type of ML in which model is trained on data that is labelled to make predictions about new data.
Training Dataset: collection of data used to train a ML algorithm. It can help teach ML models to generalize patterns from data, enabling it to make accurate predictions on new data.
Transformers: type of deep learning model that processes input data in parallel to enable the capture of complex relationships and dependencies in sequential data.
Transparent AI: concept of making AI systems more understandable and explainable to humans. It is important in fostering trust, ensuring accountability, and addressing any ethical concerns with AI technology
Unsupervised Learning: type of ML where model is given data without explicit instructions on what to do with it. The goal is for the algorithm to explore the data, recognize patterns, relationships, or structures within the data.
Utterance: statement made by user (either speech or text). Utterances are analyzed by chatbots to determine the intent.