Skip to content

Latest commit

 

History

History

setup

README.md

README - Getting Started with Verseagility

This part of the documentation helps you to set up Verseagility in our own subscription and with your data, step by step. Each of the points below will lead you to the respective documentation page. At the bottom of each page you will find a link to the next page.

Follow these steps to ramp up with Verseagility:

  1. Setup of Code and Infrastructure
  2. Bring your Data
  3. Project Setup
  4. Data Cleaning Steps
  5. Training of Models
  6. Deployment (Inferencing)

Overview

image.PNG There are three main components within Verseagility:

  1. Data processing pipeline
  2. Task modeling
  3. Deployment (Inferencing)

Data Processing Pipeline

The NLP pipeline will standardize the data into a extenible JSON format, from which data preparation and training steps are initiated. These result in custom models for the pre-defined tasks, which can be deployed in a single container instance on ACI or AKS.

Supported document types: PDF, Word Document, .txt, PowerPoint, ...

Task Modeling

image.PNG The starting point for each project with Verseagility, is the project configuration (/project/\*.config.json). The configuration defines what tasks need to be addressed using the NLP toolkit.

Supported task types

  • Text/Document classification
  • Named Entity Recognition
  • Question Answering

Example use cases:

  • Automated support ticket processing (answer suggestion, automatic routing)
  • Automated ERP/CRM workflows (extraction of entities from text to auto-complete fields)
  • Chatbot extension, with custom classification and entity extraction

Supported languages:
Versegility was developed to be as language agnostic as possible. This means that the most common languages are supported (and have been tested) out of the box, while new languages require only minor adjustments to the toolkit.

Validated languages include:

Shortcut Language
en English (US)
de German
fr French
es Spanish
it Italian

Further languages are supported with multi-language pre-trained models or required minor additions to the pre-processing and pre-trained model loading. See the FAQ for details.

How you get started is described in the first page of the documentation, Setup of Code and Infrastructure