External keywords
External keywords help to define what you need to extract. For example, when searching for invoice number, you can use the actual words "Invoice" and "Invoice No." as features, because the content after them is highly likely to be the information you need. Whereas dictionary values can only be answers, external keywords define the context of the answers.
Use external keywords to create additional features:
- NER annotations (
CONTEXT_+field_name) for all keywords in the document text. These annotations are then used by NER-based Feature Extractors. - Additional features for all tokens which have similar to keyword tokens in context.
Keywords separated by the newline (each keyword on the separate line) should be provided in the keywords.txt file in the /input/resources/ref-data/corporate_number folder.
input
└──resources
└── ref-data
└── corporate_number
└── keywords.txt