Reading the Manual: Event Extraction as Definition Comprehension
This approach could enable end-users to easily construct and extend extraction frameworks, addressing a bottleneck in text understanding for information extraction tasks.
The paper tackles the problem of event extraction by using definitions from annotation manuals to guide models, achieving the ability to extract events under closed ontologies and generalize to unseen event types by reading new definitions.
We ask whether text understanding has progressed to where we may extract event information through incremental refinement of bleached statements derived from annotation manuals. Such a capability would allow for the trivial construction and extension of an extraction framework by intended end-users through declarations such as, "Some person was born in some location at some time." We introduce an example of a model that employs such statements, with experiments illustrating we can extract events under closed ontologies and generalize to unseen event types simply by reading new definitions.