Support for users of CCG software closed 7-27-20

["Sammons, Mark" <mssammon AT illinois.edu>]

Hi, Farhaneh.

The most up-to-date install information to use is in the Curator distribution itself (INSTALL in the main curator/ directory).  This should go into a lot of detail that may help answer some of your questions. (Thanks, by the way, for letting us know that the web page version is out of date -- I'll change it now).

But to give a high-level view, to run curator, you need one or more machines (if "more", then ideally they will have a shared directory or you will have to install curator on every one). For now I will assume you have one machine with a lot of RAM -- 20G is about the minimum for the current configuration of SRL.

Once you have installed the Curator, which includes compiling it (command "ant dist"), you will start a bunch of processes, possibly using the startServers.sh script provided. You should comment out the servers you don't need; to run SRL, you need POS, Chunker, Charniak parser, NER, and (presumably) Verb SRL.  (If you want Noun SRL as well, you will need another 6G or so of RAM). You will also need the main Curator server to run. You will also need a MongoDB database process running on the same machine.

The main Curator server is the point of contact for you to use to get the various annotations. You can use the Edison library if you are programming in Java (see http://cogcomp.cs.illinois.edu/page/software_view/Edison).  This server then contacts the requested services for the actual annotation and caches the result (checks cache before calling services). You can call the Curator service from another machine, in which case you need to use the hostname of the machine on which it is running. In our group, we have machine names like "thoreau.cs.illinois.edu".  The port is a value specified by the script that starts the process -- take a look at startServers.sh.

If you don't have a single large machine, but several smaller machines, you can run some of the components on different machines. Suppose you have another machine, "dostoevsky.cs.illinois.edu" with 8G. You could run NER on that machine, and all the other processes on thoreau (if you have enough memory on thoreau). Then you would change the values in curator/dist/configs/annotators.xml for the annotator with field name NER to have the machine name "dostoevsky.cs.illinois.edu" and the port you specify when starting NER there. 

To check which services are running, you can run "ps -ef" to show all processes, or "ps -ef | grep java" to show only java processes.  You should then see processes with the port numbers you specified in startServers.sh (or the defaults if you didn't change anything).

I hope this helps.

Regards,

Mark

---

From: illinois-ml-nlp-users-bounces AT cs.uiuc.edu [illinois-ml-nlp-users-bounces AT cs.uiuc.edu] on behalf of farhaneh farahani [farhane_farahani AT yahoo.com]
Sent: Sunday, January 26, 2014 3:43 AM
To: illinois-ml-nlp-users AT cs.uiuc.edu
Subject: [Illinois-ml-nlp-users] Problem in Using SRL system of Curator or Edison

Dear All,

I am a beginner at using Illinois Curator. I am going to do SRL on my raw data. But, I think firstly I need to install Curator or Edison. I found different guidelines, but I don't know which one I should follow. What I have followed so far is this documentation:

http://cogcomp.cs.illinois.edu/curator/INSTALL.html

And the steps were done successfully (but I do not know how I can check whether Curator is running).

But, then, here :
http://cogcomp.cs.illinois.edu/curator/servers/

It seems that there is no SRL server.

So, I do not know what is the next step. Also, I do not know what "hostname" and "port" mean in Curator. Everywhere in guidelines I find example of host names like "my.host.name". Can anybody give me a real working example? I mean assuming that I have installed Curator in X directory, what must be the "host name", or even does the hostname depends on the installation directory or the system is going to use an online server?



Best Regards,
Farahani