home | introduction | examples | features | manual | site map | contact

new features in v1.3 (Crunchy Frog)

In addition to the many features in previous versions of brat, the following are new in version v1.3 (Crunchy Frog):

Normalization annotation

brat now implements various features for the visualization and creation of annotations linking text to external lexical, ontological, and database resources.

Information popup for normalized annotation
Popup showing additional information from Wikipedia for a normalized annotation.

These features allow brat to be used to create annotation for normalization (or entity linking/grounding) tasks such as Wikification [PDF]. For details, see the brat normalization page and Stenetorp et al. 2012 [PDF].

Discontinuous annotations

all text-bound annotations can now mark discontinuous regions of text as well as continuous spans.

example annotation
Example with discontinuous annotation ("North [...] America")

This feature allows single annotations to mark arbitrary extents of text, including multi-sentence or multi-paragraph regions.

Embedded brat

several improvements have been implemented to the brat client and its API to make it easier to embed brat in web pages and other browser-based applications.

example annotation
Two embedded brat visualizations showing different layers of annotation created by the Stanford CoreNLP tools.

For more information, see the page on embedded brat.

Annotation comparisons

A number of features for supporting the comparison of multiple sets of annotations for the same documents have been introduced in brat v1.3.

comparison example
Screenshot of brat in side-by-side comparison mode showing two sets of annotations for a single document.

Parallel viewer (as shown above) can be accessed by an URL like this:

/brat/diff.xhtml?diff=/left/collection/#/right/collection/document

It shows two collections side by side, showing a matching left document for a document in the right collection. UI is only bound to the right collection. It is not very polished, so if you change the right collection (e.g. using the collection browser), you will have to manually change the left collection in the URL.

The "diff and mark" tool is used to easily note the differences between two collections (e.g. between annotator and gold standard, or between two annotators). It can be found in the tools directory, and is used like this:

$ tools/diff_and_mark.py [-f|--force] first second result

It works for most permutations of documents and collections for first, second and result, giving the most intuitive behaviour for each. It will create the third collection (or document) result, which is a copy of second with added annotations detailing the differences from first.

Tutorials within brat

To help annotators start using brat, the tool now includes hands-on tutorials.

tutorial example
Screenshot from a brat tutorial document.

Collections containing tutorials are automatically set up along with example corpus annotations on all new brat installations.

Standalone brat server

brat v1.3 includes an experimental standalone server implementation that avoids the computational costs of serving brat over CGI using a separarte web server such as apache. By preloading the brat server-side software, the standalone brat server can reduce response times by up to 90% compared to conventional use via CGI.

The standalone brat server allows is easy to set up and allows brat to run on machines that do not have a separate web server installed.

For more information, see the quick start with standalone server instructions.

Configuration improvements

Many aspects of the brat configuration have been improved in v1.3, allowing easier and more fine-grained control of both the visualization and annotation functionality.

example annotation
Example configured to show additional arrow heads with labels for dependency annotation visualization.

See the page on upgrading to v1.3 (Crunchy Frog) for a brief summary of changes in the configuration.

... and more

brat v1.3 features dozens of fixes to minor issues, optimizations, and improvements to usability and documentation. As the brat development process is fully open, like brat itself, you can see the full detail of everything that went into this release on the brat issue tracker.

Return to the brat home page to try it out!