Deploy WKS ML Annotator Model After Type System Modification

Learning Objectives

The objective of this tutorial is part 2 of a previous tutorial which is to edit the type system, continue with previous annotations, train/evaluate the ML annotator model then re-deploy the model to NLU to extract patterns, keywords from unstructured text. This use case is specific for the automotive industry but can be used for other domains as well.

Prerequisites

• IBM Cloud account: If you do not have an IBM Cloud account, you can create an account here
• Provision a Watson Knowledge Studio instance within IBM Cloud & creation of a workspace.
• Basic knowledge of Watson Knowledge Studio process workflow & creation of roles by viewing docs.

Estimated Time

It is recommended to annotate a variety of a large set of documents for greater accuracy however, estimated times are as follows:20 docs = 2 hours , 10 docs = 1 hour, 5 docs = 30 min and a small sample size will be used for this exercise.

Challenge

Customers may have a need to modify the type system that they have created after they have deployed their model to NLU as a continuous training/testing practice for greater optimization and accuracy of the model.

Modification of Type System for Continuous Training-Adding an Entity Type Person

Create New Annotation Set for Existing Annotated Documents

Create New Annotation Task to Pick up New Entity Type & Annotation Continuation

Continuation of Annotation – Add Mention of Entity Type Person

Submit All Annotations for Review & Accept as Ground Truth

Train & Evaluate

Train & Evaluate

Take Snapshot ( if pleased with results and plan on doing workspace revisions later, re-training and re-evaluation of model)

Deploy to NLU

Apply Type System Updates (Quick Method)

• After modification of type system by adding entity type, edit existing annotation task, then click apply type system updates and proceed with remainder of steps above