I don't think so. Much of the discussion on this list has focused on the strong decoupling of the type system (and other properties in the analysis engine descriptor) from the analysis engines themselves. But, what are the circumstances under which a suitable type couldn't be described ahead of time? Even if the type that you think you need to create "at runtime" is elaborate and somehow generated from something in the documents that are being processed, it should be still possible to define some recursively defined annotation type that would cover all the cases, or am I wrong? In the simplest case, where all the generated annotations are flat, you could simply create an annotation type that has a string feature and set the value of the feature on the fly.
-----Original Message----- From: Ahmed Ragheb [mailto:[EMAIL PROTECTED] Sent: Sunday, November 04, 2007 9:32 AM To: [email protected] Subject: Can annotators dynamically set the typesystem it will create at runtime Dear all, For some application, the annotator needs to create an annotation of a type that is not defined in the type system. It will be consumed a component that doesn't need to know the type system in order to process the annotations. Is it possible that the annotator dynamically set the typesystem it will create at runtime? Thanks Best Regards, Ahmed ------------------------------------------------------------------------ ------- Ahmed M. Ragheb Software Engineer Advanced Technologies Group IBM Cairo Technology Development Center IBM WTC, Egypt Branch, P.O.B. 166, El-Ahram, Giza, Egypt. Tel: DID: (202) 35361048 S/B: (202) 35362536 Ext: 1048 e-mail: [EMAIL PROTECTED] "We think in generalities, but we live in details.", Alfred North Whitehead
