Machine Learning

Alfred's Machine Learning capabilities allow users to tailor file processing pipelines by training and deploying custom classification models specific to their needs. Initially, Alfred uses zero-shot and few-shot classification strategies, leveraging large language models (LLMs) for general classifications and extractions. As users provide classification feedback and more examples accumulate, tasks transition to internally trained models for improved accuracy and domain-specific performance.

This streamlined process enables users to handle training and deployment directly within Alfred's Web Console, eliminating the need for external software. Additionally, external models can be integrated using webhooks.

Accessing the Machine Learning Screens

Alfred's machine learning module has two main screens that can be accessed from any other screen within the Web Console. In order to access these screens the user must follow these steps:

1. From any screen hover the mouse pointer over the username located on the top-right section of the screen.

1. In the dropdown menu click on your desired option of either Training or External Classifiers

Training

The Training Screen contains all of the necessary tools for the user to setup, train and evaluate a custom classification model using external data. More information about this screen, including the data format needed for training, can be found in the following page:

Training

External Classifiers

The External classifiers screen contains all of the necessary tools to allow users to connect Alfred to an externally hosted Classification Model. More information about this screen can be found in the following page:

External Classifiers