Back to Multimodal Annotation Tools Workshops site.
Version 2 Format (and sample code)
This version includes many changes suggested by the community.
I have moved to the tier metadata structure suggestd by Thomas Schmidt because it made a whole lot more sense than my initial effort.
When moving to ELAN, I dropped the AG-default naming convention to be in accord with XML namespace conventions (e.g., "ts191" instead of "from-elan:AG1:ts191").
Corrected the ELAN-specific metadata as specified by Hans Sloetjes.
Corrected bug where the converter would not correctly read in an AG file without the Tier metadata. It will now happily put all the annotations on it's own default tier.
Removed empty "signals" attributes
changed startAnchor and endAnchor to start and end to match the ag-1.1 dtd
elan2ag-src2.tar.gz - the source code and binary and necessary libraries for doing the conversion from ELAN to AG and back. The README has basic instructions.
roundtrip.tar2.gz - the before, mid, and after files as they were run through elan2ag. elan-example3.eaf is the original, to_ag.xml is the AG file, and back_to_elan.eaf is after conversion back to ELAN. Also in this version is an AG file that came from exmaralda. So there's an example of exmaralda -> AG -> ELAN. ag_out.xml is the AG file from exmaralda and exmaralda_ag_to_elan.eaf is the resulting ELAN file.
Although the above files should contain everything a developer needs to work with them, as of 30 March 2007, Jeff also writes:
To help other developers convert their annotations to/from AGs, we are providing sample Java code which converts from Elan to AG and back again. This code is based on the Annotation Graph Toolkit (AGTK) Java code.
This code is provided by Jeff Hoyt of MITRE.
NOTE - This proposed format is just that - proposed. We welcome feedback on modifying it to suit your particular needs. Please provide feedback via the listserv set up for this purpose (invitations sent to this listserv under separate cover).
elan2agWriteup.doc documents the process and assumptions made during the process. Includes a description of the code for elan2ag.
elan2ag-src.tar.gz - the source code and binary and necessary libraries for doing the conversion from ELAN to AG and back. The README has basic instructions.
agtk-src.tar.gz - the source code and binary of the AGTK. This is for reference only - the jar file is also included in the elan2ag-src tarball.
roundtrip.tar.gz - the before, mid, and after files as they were run through elan2ag. elan-example3.eaf is the original, to_ag.xml is the AG file, and back_to_elan.eaf is after conversion back to ELAN.
Although the above files should contain everything a developer needs to work with them, as of 30 March 2007, Jeff also writes:
"The latest elan2ag and agtk stuff is currently up to date and available via CVS from the AGTK project on sourceforge (http://sourceforge.net/projects/agtk) - anyone can check that code out (see http://sourceforge.net/cvs/?group_id=25216 for instructions). I'll do the actual releases soon (probably tomorrow night), which will include the -src tarballs attached as well as -bin verisions that don't have the source code."
Back to Multimodal Annotation Tools Workshops site.