A new architecture
has been introduced that makes GrammarScope a 4-tier application.
The server and server-glue tiers can be operated locally (within the browser's JVM) or remotely on a server (on a distinct JVM).
new level of abstraction
A level of abstraction (the "artifacts" level) has been introduced
that makes the GrammarScope browser independent from the Stanford libraries.
GrammarScope can handle data from both the Stanford Parser or the CoreNlp pipeline.
Named entities and coreference are now displayed.
The CoreNlp pipeline has been extended to handle grammatical structures.
The display modes have been extended to include:
- typed dependency annotation (the rendering is similar to the one brat offers)
- dependent annotation (only dependent subtree is marked)
- named entity and coreference annotations
- text highlighting
- tooltips with
- grammatical structures
- typed dependencies
- extended data: coreference chain, named entities
- reference data (definition, samples)
- text mode information
All 7 modes of typed dependencies can be selected (ccprocessed, collapsed, ...)
A subject-predicate-object graph is available (after coreference resolution).
The JUNG library has been extended to produce enhanced graphs.
GrammarScope can operate as client to JEE application servers. JBoss and GlassFish servers are supported. The result data is serialized.
GrammarScope can operate as client to a Tomcat server. The result data is transmitted in XML form.
reads XML data
GrammarScope can read XML files in both Stanford pipeline and extended flavours.
The editor can generate modifications of the grammatical relation model on the fly through java class instrumentation (using Javassist)