For appeals, questions and feedback about Oracle Forums, please email oracle-forums-moderators_us@oracle.com. Technical questions should be asked in the appropriate category. Thank you!
Interested in getting your voice heard by members of the Developer Marketing team at Oracle? Check out this post for AppDev or this post for AI focus group information.
Hi Team,
Key values pairs are not completed extracted for an invoice PDF using AI document understanding services (pre-trained model). can I get some help to see how to troubleshoot this issue ?
Thanks, Sreekanth
Hi
I’ve successfully implemented a metadata design pattern to execute and store questions that may change over time to form a generic platform.
The first entity represents the questionnaire type ( A, B, C or whatever). The second entity represents the questions. For each question, stored as an inference table in Excel, provide the questionnaire text, the question answer field for which you will need several attributes ( number, Boolean, date etc). Add further attributes for whether the answer is required or not, and to handle any list of values you have for a given attribute, with of course an identifier.
I haven’t always modelled as direct parent-child as many times questions are needed in multiple surveys. Thus, a model with indicator flags to allow the use of a question in any questionnaire is more useful.
Create a relationship for each questionnaire type ( A, B, C etc) and their questions. Display the entity in a Screen. Show only the necessary answer field ( show Boolean answer if type is Boolean and so on) and render it required if the required metadata value is true.
Output the inferred questions with their answers to a custom generic object.
Whilst having a few bits of overhead due to generic structure having to hold answer columns that don’t get used in every question it is easy to build and maintain.
I have successfully implemented this for integrations with Oracle B2C Service, Siebel, Salesforce and other solutions to provide a questionnaire system with further adaptation including attachments, child and grandchild questions, images and more. What you can do will to some degree be dependent on limitations of the connector / the ability to map data to custom tables and so on.
Certainly it can be done with this approach to store the metadata in Excel and infer the instances that serve as both question and answer storage. Excel evidently has the advantage of being accessible and maintainable and copyable - many customers want the possibility to clone existing complex questionnaires in minutes not hours. Part of this model can be implemented in a Decision Service also with the same effect.
Hope that helps. If you have specific questions go ahead. Much of the above was originally modelled on how Siebel stores and manages surveys, for information.
I haven’t gone in to great detail because this is not the purpose of this thread.
Richard