Integrate DeepSRL into GATE
This is a java wrapper to execute the DeepSRL approach, parse the STDOUT and add annotations to GATE as a Processing Resource.
It does not bundle DeepSRL, as it has its own license.
-
Follow the instructions on DeepSRL to fulfill the listed prerequisites. (The nltk and tcsh prerequisites are not needed)
-
Extract the models
conll05_model.tar.gz
andconll05_propid_model.tar.gz
-
Place the packaged script
gate_deepSRL.py
(seesrc\main\python\
) into the DeepSRL folderdeep_srl-master\python
DeepSRLAdapter starts an own python process and handles communication via Stdin/Stdout.
These are the parameters which have to be set when instantiating the DeepSRLAdapter as a processing resource in GATE.
-
deepSRLExecutable
: Path to thegate_deepSRL.py
-script (inside thedeep_srl-master\python
-folder, see above) -
modelPath
: Path to the extractedconll05_model
-folder -
propidModelPath
: Path to the extractedconll05_propid_model
-folder -
pythonExecutable
: Path to the Python2 executable -
printOutput
: Print Stdout while script startup -
printError
: Print Stderr while script running
DeepSRLClient communicates with an server via TCP/IP. You have to start the gate_deepSRL.py
-script yourself with the --server
parameter and optionally --host hostname
(default: all available network interfaces) and --port port
(default: 6756). The communication is not encrypted.
These are the parameters which have to be set when instantiating the DeepSRLClient as a processing resource in GATE.
-
host
: Hostname of the server (default: localhost) -
port
: Port of the server (default: 6756)