Web interface to manually annotate named entity mentions in newspaper articles with the correct DBpedia link(s), if any. Produces labeled data sets for training and evaluating the DAC Entity Linker.
To start the web applicaton, run:
./web.py
This will start a Bottle web server listening on http://localhost:5001
. To work on a specific data set, add the set name, e.g. tve
, to the URL:
http://localhost:5001/tve
The training application will automatically show the first example from the requested set that hasn't been labeled yet.
In order to request a specific example from a data set, use either id
(unique identifier) or index
(index number within the set) as a parameter:
http://localhost:5001/tve?index=1
Selecting one or more candidates from the user interface as the correct links and navigating to the next example using the menu in the upper right corner will save the selection.
Adding a new article to the set:
http://localhost:5001/tve/edit?action=add&url=http://resolver.kb.nl/resolve?urn=ddd:010734861:mpeg21:a0002:ocr
Adding a specific named entity to the set:
http://localhost:5001/tve/edit?action=add&url=http://resolver.kb.nl/resolve?urn=ddd:010734861:mpeg21:a0002:ocr&ne=Einstein
Removing an article from the set:
http://localhost:5001/tve/edit?action=delete&url=http://resolver.kb.nl/resolve?urn=ddd:010734861:mpeg21:a0002:ocr
All datasets are stored as an art.json
file in a folder with the dataset name within the users
folder. To create a new, empty dataset named foo
:
$ mkdir users/foo
$ echo '{"instances": []}' > users/foo/art.json
The art.json
file has to be writeable by the web application and is expected to contain at least an empty list of instances.