ASR stands for Automated Speech Recognition. It is a technology used in many digital assistants to recognize speech. In contrast, Speech-to-Text allows recognizing voices.

Avatars are digital representations of a user or a machine. The embodiment is in the foreground here. They can have different degrees of intelligence and take on different forms of representation. Photorealistic avatars - as numerous HCI studies show - set higher expectations for the intelligence of the machine. Therefore, one should choose carefully between a comic-like, realistic avatar or a mixed form.

They recently caught attention with the metaverse worlds such as the Sandbox or the RTFKT  creators-led company bought by Nike.

Check out our blog on: Great tools to develop avatars in the metaverse.

A bot (from "robots")  is a software to automate tasks. The term bot is often referred to malware or cybersecurity software to mitigate fraud or spamming.  Via CAPTCHA's and honeypods you can reduce unintended bot actions on your applications or websites.

Types of bots:

• Socialbots: Bots are also used in social media (e.g. instragram) to create automated followers and for non-organic increase of a leads, followers or customers. 
• Spambots, which use accounts created by registration bots (e.g. via honeypods) to send spam emails to addresses collected by harvesters.
• DDoS bots, which are computer worms that perform joint attacks such as DDoS attacks on command. Corresponding botnets sometimes consist of hundreds of thousands of computers.
• Social bots, which increase the number of hits on certain pages or make content appear in a better light through fictitious ratings, comments or reactions

Check out our article on how create automated communication on instagram. 

A chatbot is derived from "chatter" and "bot" and is software that typically communicates proactively with the user and can - but does not have to - perform tasks for the user by executing automated actions in back-end systems. 

There are different types of chatbots:

• Rule-based chatbots (e.g., ELIZA): they provide predefined responses and establish communication paths.
• Clickbots: the user is guided through the conversation via buttons and rich elements. For example, as a user, you can click on various buttons (also called "chips" or "quick replies") to move to the next bot response.
• AI bots: AI bots include bots that are based on natural language processing (NLP), which means they learn to understand the user's input through automatic or manual training of utterances.
• Hybrid bots: Some bots are a mix of rule-based and AI bots.

A  large language model based on GPT-3 that adds a conversational layer on top of it to remember the context and react to prompt engineering.

A conversational interface is an interfaces that communicates like a human being in a message-based and back and forth way.

Conversational Interfaces include chatbots, voicebots.

Chitchat is referred to as "small talk" with chatbots. Chitchat means "informal conversation  where the topic doesn't matter much". 

Most chatbot tools and designers incorporate chitchat such as "How are you doing today?" or predefined response phrases in case the user asks the bot questions such as "How old are you?" or "Where are you from?". 

The advantage of incorporating small talk elements is that they allow for some natural communication, just as happens between real people when they meet for the first time or want to start any kind of conversation (e.g. negotiation, date). 

Conversational AI refers to a field in artificial intelligence which has the goal to create "conversational" applications with the use of automation and natural language processing. 

Conversational AI is not equal chatbot but chatbots as they may be rule-based and programmatically built. Chatbots are often  based on conversational AI platforms or algorithms.

As described on moveworks a copilot is "an AI conversational interface that uses large language models (LLMs) to support users in various tasks and decision-making processes across multiple domains within an enterprise environment. 

An enterprise copilot is a fluid conversational interface that connects your employees with every business system."

An Entity is a key element in a text. For example "Apple" is a type of the entity "fruit". So entities could be names, people, places, etc.

According to Microsoft's learning site few shot prompting is:

A common way to adapt language models to new tasks is to use few-shot learning. In few-shot learning, a set of training examples is provided as part of the prompt to give additional context to the model.

When using the Chat Completions API, a series of messages between the User and Assistant (written in the new prompt format), can serve as examples for few-shot learning. These examples can be used to prime the model to respond in a certain way, emulate particular behaviors, and seed answers to common questions.

One example from DLAI is: 

prompt = f"""

Your task is to answer in a consistent style.

<child>: Teach me about patience.

<grandparent>: The river that carves the deepest \ 

valley flows from a modest spring; the \ 

grandest symphony originates from a single note; \ 

the most intricate tapestry begins with a solitary thread.

<child>: Teach me about resilience.

"""

response = get_completion(prompt)

print(response)

Wikipedia: Subtask of information extraction that seeks to locate and classify named entities mentioned in unstructured text into pre-defined categories such as person names, organizations, locations, medical codes, time expressions, quantities, monetary values, percentages, etc.

Huge pre-trained models that use deep learning and neural networks. 

Examples: GPT-3 from OpenAI, PaLM or LaMDA from Google, OPT from Meta, Megatron-Turing from Nvidia/Microsoft, Jurassic-1 from AI21 Labs—

They are autoregressive, self-supervised, pre-trained, and based on so-called Transformers.

Rule-based chatbots consist of a tree-like flow and have predefined answer sets from which they select their answers. This means that they do not generate "new" answers that have never existed before, which would be the case with "generative AI models". Find out more on generative AI models here: 

Consequently, the user is guided in the front-end with follow-up questions or via rich elements such as Quick Replies. The advantage of such chatbots is that they provide a high level of control over the conversation. The disadvantage is that the experience can seem static and robotic. 

Natural Language Processing (NLP) is the process of Natural Language Understanding (NLU) and Natural Language Generation (NLG), or:

NLP = NLU + NLG

NLP is a intersection between Linguistics and Artificial Intelligence.

It takes advantage of different machine learning algorithms, methods and models by "processing language"  to understand written or spoken language and text the same way as human beings can. 

Natural Language Generation is a part of Natural Language Processing (NLP) and includes the generation of language in written or spoken form.

A typical Natural Language Generation service is Polly from AWS.

 “Rachel Wood” Twitter feed, the term 'prompt drift' refers to the alterations in the responses generated by different versions of the same model when given an identical input or prompt. This drift is a consequence of the modifications made by developers in the model's iterations. 

In a non-technical way, a "Prompt Pattern" is like a recipe for asking questions to AI models like ChatGPT. It's a standardized way to structure your questions (prompts) to get more accurate and useful answers. These patterns help you go beyond simple queries and enable you to perform complex tasks with the AI. They're especially useful when you want consistent and specific outputs.

Natural Language Understanding includes the interpretation of language through a NLU module in Natural Language Processing.

Stemming is used in Natural Language Processing to accelerate interpretation of words by reducing them to their word stem. 

An utterance is a sentence or snippet of text which is fed to an NLP model (e.g. NER) to train a chatbot to understand intents of the user.