Getting Started
Download and build the toolkit yourself from the source.
Directory structure
com.ibm.streamsx.nlp
- the toolkit directory containing the operators, functions and librariessamples
- Sample projects demonstrating the usage of each toolkit operator and functionsetup
- File for the installation procedure of required modulestest
- Test suits for regression and performance tests.
Building the toolkit
The toolkit needs to be build before you can use it in a SPL application.
To build the toolkit libraries, cd
to the com.ibm.streamsx.nlp
folder and run the make command:
make all
Setup
The toolkit operators ContentClassification and LinearClassification require Python 2.7 be installed.
To install the required and verified extra modules, cd
to the setup
folder and run the install command:
(sudo) pip install -r requirements_2.7.txt
In the IBM Streams 4.2 Quick Start Edition VM, cd
to the setup
folder and run the install command:
./setup_streamsV4.2-qse.sh
Running the sample applications
To build and launch a sample application, cd
to the samples/<SAMPLE_APP>
folder and run the commands:
make all
output/bin/standalone
<SAMPLE_APP> is one of the sub-directories of the samples folder, e.g. UimaTextSample.
Generate the documentation
Call make make-doc
in the com.ibm.streamsx.nlp
directory to generate the toolkit documentation pages.
Open the index.html
file with your browser in the com.ibm.streamsx.nlp/doc/spldoc/html
folder to view the toolkit reference documentation.
Running the test cases
Call ./runtests.py -f alltests.txt
in the test
directory to perform the regression test.
This compiles and launches each sample application and verifies the resulting output.
At the end of the test run, the result is printed to the terminal output.