Unable to map the entity when it comes on different segments. For Example We need to map a tabular data, the field claim number is coming in the first segment and the value is coming in the second segment . Watson splits the single row in to multiple segments. Since the claim number and claim value are coming in different segments , we are unable to map the relation type between the header (Claim Number) and the value (Claim Number Value). Is it possible to override the segmentation(or paragraph split) capability of IBM WKS ?
And explore the tokenization process for overcome the issue . Since the values coming are not predective, we are unable to create the machine /dictionary tokenization property.Please find the document for referance.
I think tabular data means data in table format. Generally, ML model of WKS aims at handling with natural language, not structured data such as data in table format. In this case, ML model might work, but does not utilize "context" of the data enough.
It's true that WKS has Dictionary-based Tokenizer to control sentence split (typically, for example, title). However, your sentence split at "Changed." in your case, and it seems natural.