We are trying to develop some application which will respond to queries of the user and the complete data will be annotated inside Watson Knowledge Studio. Queries will be answered according to that data.
As User is a human here so there can be typos in the query strings. As data in the Watson Knowledge Studio have correct data without any typos. Is there some way to handle typos in Knowledge Studio itself?
Another thing which we were thinking is about Conversation API. As it can handle typos and learns from it for the next queries. So can we combine these two things where Conversation API will be the first layer and will pick what to respond from Knowledge Studio
Answer by MitchMason (5376) | Apr 13, 2017 at 12:27 AM
Knowledge studio does entity extraction. It should do okay with typos (with enough training data) Conversation has intents and entities. The intents can cover for typos as they are building a statistical model and then finding the closest match. Entities in conversation are (currently) string matches so it won't help for entities specifically.
What kind of application are you making and maybe I can try to recommend a solution?
I want to extract the name of the Product and Feature from the query. In knowledge studio, we will map the products and features. Corresponding to every product and feature, we have some data to show. This is the actual thing I want to do. Can you recommend something for this.
Another thing, I want to ask apart from this application that can we use both conversation service and Knowledge Studio together?
Answer by MJonker (537) | Oct 18, 2017 at 03:09 AM
As far as I understand, there are NO API's to access Knowledge Studio models directly. You can do this:
Deploy a machine-learning or rule-based model to use with the Watson Discovery service.
Deploy a machine-learning or rule-based model to use with the Natural Language Understanding service.
The way to integrate WKS with Conversation Service, would be via the Discovery Service.