Marc Dymetman, Aurelien Max, Kenji Yamada
ACL 2003, 41st Annual mtg of the association for Computational Linguistics, Sapporo, Japan, July 7-12, 20030.
Several text-processing tasks, such as answering the emails sent to a company by its customers (to take just
one example among many), require a reliable understanding of the content of incoming messages. This
understanding can only be done currently by humans, and represents the main bottleneck to a complete
automation of the processing chain: other aspects could be delegated to such procedures as database
requests and text generation. Current technology in Natural Language Understanding or in Information
Extraction is not at a stage where the understanding task can be accomplished reliably without human
intervention. In this position/programmatic paper, we advocate an interactive approach where: (1) the building
of the semantic representation is under the control of a human author; (2) in order to build the semantic
representation, the author interacts with an intuitive textual interface to that representation (obtained from it
through an NLG process), where some "active" regions of the text are associated with menus that display a
number of semantic choices for incrementing the representation; (3) the raw input text to be analyzed serves
as a source of information to the authoring sytem and permits to associate likehood levels to the various
authoring choices; the choices are then ranked in each menu according to their likehood, allowing a speedier
selection by the author; when the likehood of a choice exceeds a certain threshold, this choice can be
performed automatically by the system (but in a way that remains revisable by the author). The system acts
as a flexible understanding aid to the human operator: bu tuning the threshold at a low level, it can be used
as a purely automatic, but somewhat unreliable, information extraction or understanding system; by tuning
the threshold higher, it can be used as a powerful interactive guide to building a semantic interpretation, with
the advantage of a plain textual interface to that representation that is easily accessible to general users.
Report number: