SkySync is now DryvIQ | Learn More

Enhance accuracy of classification via calibration blankword

The calibration module is designed to leverage out of the box machine learning and artificial intelligence model entity types and customize them or “train them” to be more accurate at finding company specific and proprietary data.

While the entities were designed to find and classify a broad range of industry standard data—such as government forms and formatted documents—the calibration module enables the ability to train the models with company specific forms and documents. The objective is to get closer to the business, by refining the artificial intelligence to look for company specific data points.

A repository of content leveraged by the calibration and verification module to train and test the different artificial intelligence models:

  • Two modes of population—user uploaded files
  • Two modes of population—items found during scan [disputed items] - coming summer 2022
  • Ability to organize samples of content and files
  • Content in the repository can be used across all models
  • Upload only one copy of a file if it is needed to train different models

Used to customize each of the artificial intelligence models to meet company specific data classification needs.

The training module is the “heart” of the operation and used to build out entities that are relevant to your business:

  • Pre-trained A.I. models available to be customized for companies needs
  • Guided process will direct the user to provide the required information to train the individual model
  • Samples from the repository will be used for training the models—models can be trained as often as needed
  • Newly trained versions of the models can be selected to install into the platform

Designed to help test out the accuracy of the customized A.I. models and make educated decisions on which model to apply to the platform.

  • Ability to test trained models using test samples from the samples repository
  • Testing a model will show how a selected file will be classified by the model
  • Match accuracy is displayed for each file tested and classified
  • Used to decide on which version of the model should be used in the platform