/nwr-english-pipeline-vm

Instructions to download and set up the NewsReader pipeline for English

Introduction

This is a virtual machine (VM) with all the required modules and NLP processors to run the English pipeline developed within the Newsreader project for event extraction. The pipeline expects a NAF document as input (see for example docs/input/example.naf) and produces NAF documents with the annotations of the different linguistic processors.

Pre-requisites

The are some pre-requisites the host machine has to fulfil for running the VM inside it:

  • Linux operating system (any recent flavor would do it)

  • In-kernel KVM virtualization capabilities. You can also determine if your system processor supports KVM by running the following command:

      % grep -E 'vmx|svm' /proc/cpuinfo

    if this command returns output, then your system supports KVM. You also have to verify that the KVM-related feature is enabled in the machine's BIOS.

  • 64 bit CPU (x86_64)

Download the virtual machine

You need two files for running the VM.

The first is the VM image with all the required modules installed. The second is an short XML document which is needed for running the VM.

Running the VM

Preliminary steps

For running the VM, you need first to do some preliminary steps:

  • Install the necessary software into the host machine. On Deban/Ubuntu machines this includes the following packages:

    • qemu-kvm
    • libvirt-bin
    • virt-manager

  • Tweak the the XML file:

    • Put the absolute path to the IMG file above in the file attribute of <source> element (under <disk>, around line 22)
    • Currently the VM is configured to use 12Gb RAM. You can change this value by editing around line 4 (<memory> and <currentMemory> elements).
    • If you experience problems running the VM, maybe you need to change the <emulator> element (around line 19) and put the correct path of the kvm emulator executable in your system.

Running the VM

From the host machine, cd to where the IMG and XML documents are and run the following:

% virsh create newsreaderpublicvm_v3.xml
Domain newsreaderpublicvm_v3 created from newsreaderpublicvm_v3.xml

The machine should be running now. You can test this using the virsh list command:

% virsh list
 Id    Name                           State
----------------------------------------------------
 2     newsreaderpublicvm_v3          running

Note: The VM needs some time to completely load all the modules and services, so please be patient until the login screen appears.

The IP address of the machine is fixed: 192.168.122.92

Now you can connect to the VM.

Connecting via ssh

As said before, the IP address of the VM is 192.168.122.92. Thus, being on the host computer, just ssh to the VM using this address and the newsreader user:

% ssh newsreader@192.168.122.92

The password is readNEWS89

Running the pipeline

Inside the VM, you can run the NewsReader pipeline by executing the following command:

$ pipeline/run_pipeline.sh docs/input/example.naf > docs/output/example.naf

If you want to analyze the debug messages of the NLP modules, run this other script instead:

$ pipeline/run_pipeline_v3_debug.sh docs/input/example.naf > docs/output/example.naf

Shut down the VM

Logout from VM and then, from the host computer:

% virsh destroy newsreaderpublicvm_v3

Alternatively, and being on the VM, run the following command:

$ sudo shutdown -h now

List of files and directories in th VM

+ README.md      this file
+ components/    the different NLP processors
+ pipeline/      scripts for running the pipeline
+ docs/          directory with documents

Pipeline components

The current version of the English pipeline executes the following NLP processors following this specific order:

components/EHU-tok.v30
components/EHU-topic.v30
components/EHU-pos.v30
components/EHU-wikify.v30
components/EHU-nerc.v30
components/EHU-parse.v30
components/EHU-corefgraph.v30
components/VUA-opinion-miner.v30
components/EHU-ukb.v30
components/VUA-ims-wsd.v30
components/EHU-ned.v30
components/VUA-popen-nedreranker.v30
components/EHU-srl.v30
components/VUA-srl-postprocess.v30
components/FBK-time.v30
components/VUA-eventcoref.v30
components/FBK-temprel.v30
components/FBK-causalrel.v30
components/VUA-factuality.v30