BAT Status

home > status

Overview of main functionality

extent annotation	+
extent adjudication	+	minor bugs in how the default value for judges is calculated
attribute annotation	+
attributes adjudication	+
relation annotation	+	first version, limited to only one attribute
relation adjudication	+	see note above

Recent Changes

July 28th 2010

1. Added propagation of annotator data for all layer types. Now the judgements of one annotator (typically from a file upload) can be propagated to all other annotators on a file-by-file basis.
2. Fixed a new bug in layer data upload where assignments would not be properly updated.
3. Added setting to layer data upload which lets the administrator decide what the assignment status is of the annotator of the newly uploaded data.
4. Fixed broken link in page for administrator source data upload.
5. Finished incomplete and incorrect help on the administrator's corpus export page.

July 26rd 2010

1. Adjudication phase for relations was added as well as functionality to allow the administrator to view relations.
2. Administrators can now set and change passwords of annotators.
3. Added first version of database consistency checking script.
4. Fixed small bug in deletion of layers and corpora.
5. Manual Updates to the manual.

July 23rd 2010

1. Added export and upload functionality for relation layers.
2. Made relation code independent of TimeML.
3. Added log viewer.
4. Added relation layers statistics to the administrator's layers page.
5. Updated help entry for layer data upload page.
6. Some, as of yet incomplete, changes to allow proper relation annotation for judges.

February 26th 2010

1. Added first version of proper relation annotation for annotators (but not yet for judges and administrators).
2. Optimized code for inter-annotator agreement on attributes
3. Fixed bug in the way that judge attributes overwrite attributes given by annotators.
4. Fixed code that required elements in a sequence to have consecutive lex_ids, but that was not always the case, especially for uploaded data.
5. Added script for fixing duplicate tag id problems with chinese corpus.
6. Made sure that the extent in the attribute box is printed with the tag id, useful for long sentence, especially if the same string occurs as two events.

January 20th 2010

1. Added IAA pages back in, but embedded in a more general page with some extra prose.
2. Extent layers in sequence mode now display and update the number of instances correctly.
3. Extent sequences longer than one are now properly displayed; in some rare cases, they were displayed as a sequence of single tokens.

January 7th 2010

1. Added the beginnings of a new help function for admin pages
2. Streamlined layer data upload
3. Increased allowed file size for token upload to 1Mb

January 2nd 2010

1. Changed the way sequence extents are displayed. Now to change the extent, you do not click one of the brackets around the extent, but you click the first or last element. The extehnt turns red, move the cursor around and click again. If you want to delete the sequence, just click delete when the sequence is red. The delete button is now gone.
2. Changed the way layers are displayed to the admin
3. For tokens, all extent features of all candidates were shown. Now only show those that were selected.
4. Fixed bug in display of extent layers that had another layer in the domain. It should just display the extents selected by the judge, but it would display the union of all extents selected by all annotators.

November 18th 2009

1. Corpus upload failed to properly set assignment status. Fixed.
2. Viewing attributes where extents could be sequences was buggy. This is now mostly fixed.

November 16th 2009

1. Added code to allow upload of attribute layer data.
2. Made sure that PHP sends out the right header for character encoding. By default, it would use UTF-8, overwriting the character encoding specified for the corpus.
3. Fixed more bugs in the extent layer upload code.

November 5th 2009

1. Fixed the merge scripts. It confused the tagids of events from diffent layers. I am checking on whether this had any impact on recent work.
2. Fixed an error in the extent layer upload code.

October 25th 2009

1. Bug fixes.
2. Added code to upload entire corpora
3. Code cleanup.

September 10th 2009

1. Export code used to choke on a corpus with 97K tokens. Reduced memory use of system so that said corpus could be exported. it is not clear what the new ceiling is, but it is expected to be around 250k tokens.
2. Added the support_domain feature to the layers. Elements from the support_domain of attribute layers show up in dark red, with a tag id.

August 20th 2009

1. Added button to layers page to allow deleting layer data (but not the layer itself and the assignments associated with it).
2. Added admin functionality to change corpus encoding and corpus description.
3. Fixed silly bug where some filenames were disallowed for obscure reasons.
4. Updated the administrator manual with a section on Tempeval annotation and updated the annotator manual with a section on timex3 attributes.

August 12th 2009

1. Added code to allow judging phase for attribute annotation.
2. Updated the statistics page. Works now correctly for extents and arbitrary numbers of annotators. Added Kappa scores.
3. Some more optimizations added so we can work with larger files. However, very large files with layers with many annotators may still choke.
4. Fixed bug to prevent loss of data in some situations where an annotator had multiple windows open.
5. Updated documentation.
6. Corpus export now exports all data, including layer definitions. From now on, an archive with an exported corpus contains sufficient data to recreate the corpus.
7. Added encoding as a feature to the corpus creation form. This encoding is used when corpus files are displayed.

July 24th 2009

1. Added functionality to delete a layer with all associated data.
2. Added navigation panes to the extent and attribute views, allowing the annotator to quickly go to any sentence in the document. All sentences are now numbered.
3. Extent features for sequence annotation are now properly displayed.
4. Export of larger corpora now works.