I love Watson Knowledge Studio (WKS) which has been my go-to tool to help me extract domain specific content from my documents, for different projects.
I learned that WKS is now available on IBM Cloud. That’s odd; I thought it was always a cloud service. Digging deeper, I found out that WKS is now available on the IBM Cloud (formerly called IBM Bluemix) and will be retired on IBM Marketplace. So, what has changed with WKS on IBM Cloud? It has a nicer user experience, a new administrative dashboard and I can now pre-annotate my documents with Natural Language Understanding APIs.
I also found out that I would have to move my existing projects on the WKS version on Marketplace to IBM Cloud. The sooner I switch the better it is for me since the free instances on Marketplace will eventually be removed.
You guessed right! The first question that popped up in my mind was about the effort for migrating to IBM Cloud. And here is some good news; the underlying workflow of building a machine learning model has not changed with WKS on IBM Cloud and hence the migration process is fairly simple.
What did I need to do to migrate my WKS projects to IBM Cloud?
First, I had to create a new free WKS instance on IBM Cloud. I was able to re-use my IBM ID and create an instance of Watson Knowledge Studio from the Watson Catalog with my free plan. Once I had my instance of WKS ready, migration to IBM Cloud was easy. It involved 3 steps:
1) Exporting assets (type system, dictionaries and annotated documents) from Watson Knowledge Studio on Marketplace
2) Importing them to Watson Knowledge Studio instances on IBM Cloud and
3) Building a machine learning annotator.
So, here is a quick break down for you:
1. Export assets from Watson Knowledge Studio on Marketplace
a) Export documents with ground truth
You would want to finish all your annotations and promote them to ground truth prior to exporting your documents. Only documents with ground truth are exported.
b) Export dictionaries
You can export some or all the dictionaries that are editable.
c) Export your Type System
2. Import assets into Watson Knowledge Studio on IBM Cloud
You will first need to create a workspace within your WKS instance. You can then import your resources to your workspace in the following order:
a) Import Type System
b) Import Dictionaries
c) Import documents with ground truth
Note: Projects in WKS on Marketplace are referred to as workspaces on the IBM Cloud version of WKS.
3. Build your machine learning annotator
Since you have imported your documents with ground truth, all you are left with is training your machine learning model with a click of a button. Once the machine learning annotator is built, it can be deployed to your application.
Note: When you are ready to delete your WKS project or WKS instance on Marketplace, you will need to undeploy any custom models that were deployed to either Watson Discovery or Watson Natural Language Understanding services . Otherwise you run the risk of orphaned models that are deployed in instances where you don’t want or need them!