GATE and Dialogue
GATE is being used in the Amities project to produce dialogue processing server components to run in the Galaxy Communicator architecture. This note discusses the relation between the two architectures.
Different architectures in language processing concentrate on different infrastructural problems. Galaxy Communicator concentrates on distributed processing, hooking together sets of servers and clients that collaborate to hold dialogues with human interlocutors. Data gets passed between these components as attribute/value sets, or "frames", the strucuring and semantics of which must be agreed upon on a case-by-case basis. This architectural style tends to treat components as black boxes which are developed using other tool sets.
GATE ignores the issue of distributed processing, and concentrates on facilities for bootstrapping, developing, testing and deploying language processing components. Data visualisation is an important focus of the system, as is an annotation model with associated finite state machinery and automated evaluation mechanisms. The underlying framework supports multilingual distributed data storage, I/O to many document formats, and can be easily embedded in 3rd party applications.
Sheffield have used GATE to produce a Galaxy Communicator server component, taking advantage of the GATE development environment and then using Galaxy Communicator as a communication substrate to integrate with other partners' components.
There seems a natural synergy between the two systems, GATE forming a toolset for developing servers and Galaxy Communicator tying sets of servers together to form dialogue systems.
In future work we would like to more closely integrate GATE with Galaxy Communicator, by building in support for Galaxy Communicator protocols at the GATE component level:
- Add a mode to the GATE UI to run as a Galaxy Communicator server. GATE resources would communicate with the Galaxy Communicator hub, with utterances appearing as documents in GATE. The user could then debug, send modified versions of what the system produced, etc.
- Adding dedicated dialogue document and audio-visual document support. Issues include taking input as n-best word lists or word graphs, as well as displaying audio-visual data.
- Adding a capability to bootstrap GATE-based Galaxy Communicator servers in the same way GATE will bootstrap other resources.
- Representing other Galaxy Communicator servers as resources in GATE.
This work would leverage the strengths of both systems, bringing distributed processing to GATE and langauge processing development tools to Galaxy Communicator.