Difference between revisions of "Coreference Resolver"
From Knowitall
(2 intermediate revisions by the same user not shown) | |||
Line 12: | Line 12: | ||
==Final Write-up== | ==Final Write-up== | ||
− | [ | + | Final Write-up[https://docs.google.com/open?id=0B1-38B0n59T5YjlmNDgzZWUtNWQwZS00NjgyLTk4ZDUtZDBkOWM1YjA0N2Y5] |
+ | |||
+ | ==Additional Logic Added After the Write-up== | ||
+ | * Replacement guard: If the mention matched to the argument overlaps with representative mention of that cluster, no replacement occurs. Example Fixed: argument "Japan, China, and Russia" won't get replaced by "Japan" |
Latest revision as of 21:29, 15 December 2011
Contents
Running Coref From the Command Line
Use run.sh in the coref project. Run it with no parameters to see the options usage.
Running the Web Demo
- Code under coref-webapp folder in java projects directory
- To start up, run "mvn jetty:run"
Using Coref as a Library
- Under the coref project, look at StanfordCorefResolver's main method for example usage.
- Call resolve on a document after loading in document text (can be loaded in line by line or as a string in the constructor)
- Call getResolvedExtractions to retrieve all extractions (some will be changed and some not). Take a look at ResolvedExtraction class, there is a method that allows you check if it has been changed or not.
Final Write-up
Final Write-up[1]
Additional Logic Added After the Write-up
- Replacement guard: If the mention matched to the argument overlaps with representative mention of that cluster, no replacement occurs. Example Fixed: argument "Japan, China, and Russia" won't get replaced by "Japan"