Difference between revisions of "Vulcan/ProjectPlan"
From Knowitall
(→Plan) |
(→Plan) |
||
| Line 24: | Line 24: | ||
* H1: MLN formulation will be robust for textual inference i.e., it can handle gaps in inference. | * H1: MLN formulation will be robust for textual inference i.e., it can handle gaps in inference. | ||
* H2: Open IE provides sufficient representation power and extraction accuracy levels that are suitable for inference. | * H2: Open IE provides sufficient representation power and extraction accuracy levels that are suitable for inference. | ||
| + | * H3.1: Definitions contain information that is useful for answering questions. | ||
| + | * H3.2: Pattern based extractors yield information in a format that is useful for inference. | ||
b) Identify areas for improvement. | b) Identify areas for improvement. | ||
Latest revision as of 17:40, 18 September 2013
Overview
Idea is to state where we'd expect to be in two months time.
Plan
End of November (14 weeks)
1. System goals
a) End-to-end QA system that allows batch processing.
b) KB extractors: Definitions extractor, Study guide rules extractor and general Open IE.
c) Modular implementation that allows swapping individual components.
d) Tight integration with evaluation framework allowing for easy training of system parameters.
2. Research goals
a) To test our main hypotheses.
- H1: MLN formulation will be robust for textual inference i.e., it can handle gaps in inference.
- H2: Open IE provides sufficient representation power and extraction accuracy levels that are suitable for inference.
- H3.1: Definitions contain information that is useful for answering questions.
- H3.2: Pattern based extractors yield information in a format that is useful for inference.
b) Identify areas for improvement.
