Glossary of Terms

Activity

Activities are predefined actions or instructions that can be run within other areas of Symbient AI. An Activity gives a set of instructions to an LLM Model. These instructions can be any natural language directions such as "Summarize this transcript" or "Retrieve all emails mentioned in this text."

Automation

Automations allow you to create in-depth logic systems that can perform a wide variety of functions. These Automations use individual steps called Operators, which can execute tasks such as reading or sending emails, running an Activity, passing text to an LLM Model, or reading and writing data to a webhook. These Operators are chained together to form an Automation Chain.

Menu Bar

When a user is logged in, there will be a menu on the left-hand side of the screen that allows the user to navigate between the main pages of the application.

Roles

• Company Admin: This role has access to do anything in Symbient AI. They can add/edit/delete other users and models, and even maintain their group's membership, payments, and payment methods.
• Accounting Manager: Needed for viewing Payments and Payment Methods tabs on the Settings page.
• Activities Manager: Authorized to access, add, edit, and delete Activities. Other users can use these Activities on Transcriptions, Automations, and other pages, but they are not allowed to modify or delete them.
• Actors Manager: Authorized to access, add, edit, and delete Actors. Other users can use these Actors, but they are not allowed to modify or delete them.
• Automations Configuration Manager: Authorized to access, add, edit, and delete Automation Configuration within the toolbox. Other users can use these Automation Configurations within Automations, but they are not allowed to modify or delete them.
• Automations Manager: Authorized to access, add, edit, and delete Automation.
• Data Manager: Authorized to edit and manage the process of Archiving and Deleting Transcriptions.
• Model Manager: Authorized to access, add, edit, and delete models. Other users can use these models on the Transcriptions, Automations, and other pages, but they are not allowed to modify or delete them.
• Transcription Manager: Handles access to any Transcript-related APIs as well as visibility of the Transcription and Calls dashboard in the Toolbox.
• User Manager: Authorized to add, edit, and/or delete other users from the Users page. However, they cannot change Company Admin permissions.

Model

A model in the context of Large Language Models (LLMs) is a system trained to process, understand, and generate human language. They function as a way to switch between different use cases and specializations within Symbient AI applications. There are three primary classifications of Symbient AI models, each designed for distinct operational purposes:

LLM (Large Language Model)

An LLM Model can pull information from .txt, .doc, .docx, and .pdf Data Sources and can utilize a specialized Vocabulary for better results. The primary function of an LLM is to interpret user requests and generate responses by leveraging its vast repository of text information.

ASR (Automatic Speech Recognition)

An ASR Model is used when processing Transcriptions. The main function of the ASR Model is to help Symbient recognize specific words or phrases it might not know, such as technical terms or industry-specific language, like medication names or legal jargon.

Classification

The Classification model is used to analyze user requests and identify Intents. It is particularly useful when trying to classify different types of user engagements, such as distinguishing positive from negative feedback, or identifying requests for transactions, such as placing or modifying orders.

AI Actor

An AI Actor refers to a set of additional instructions added to the base model to refine its response mechanism. These instructions are designed to tailor the model's output, ensuring it aligns with specific contextual or tonal requirements. For example, you could instruct it to:

• Be an Expert on a Specific Topic: "Answer all questions as if you're an expert on Symbient Opta and its uses."
• Maintain Specific Documents: "Your role is to ensure the user documentation for my website is always current."
• Adopt a Unique Personality: "Pretend you are a pirate in all your responses."

Intent

Intents are the underlying purpose or objective within a user's query or statement. They are used within Models to train the Slots and Input tabs. Some examples of user intent extracted from a request could be positive vs. negative customer interaction or identifying when a user wants to place or change an order.

Slots

Slots are predefined variables within a conversational model that are used to extract and process specific pieces of information from user input. These variables allow the model to understand and act upon the details of a user's request. For instance, in the context of a customer seeking to book a flight with an airline, the model could have Slots for travel parameters such as the departure city, destination, and travel date.

The data within these Slots can be extracted and transmitted to other systems for storage, verification, and initiating actions, such as triggering a travel-booking system to auto-fill a reservation. This structured approach enables the model to accurately identify and utilize the relevant information provided by the user, thereby facilitating a more efficient and tailored response to the user's needs.