CSLI/CALO - API Usage HOWTO

THIS IS Y3 DOCS...new ones coming soon

This document contains instructions to developers on what programmatic steps to take to run our CALO software and perform processing automatically in the context of the deployed MA server system. You will probably want to go through the document called "Execution Steps (Development)" first in order to work through installation issues, etc.

Classpaths, jars, and main methods

During development, the main CSLI processes (e.g. NOMOS and the Service Manager) are started using the Agent Manager. To what is happening under the hood to start these agents, please just have a look at the .sh files in the calo/run directory. These commands are run by the Agent Manager and will show you the main methods which are executed as well as the very long list of jars on the classpath. All thrid-party jars used by the CSLI system can be found as csli/*/ext/*/*.jar. Licenses and readmes are located in csli/*/ext/* for each software component. You will also need to add the CSLI project jars to your class path: csli/*/*.jar.

Ordered Steps

1. Run through the installation process described in the development docs, including setting up your calo.local.config file
2. Start up the Service Manager
   • ServiceManager csliServiceManager = ServiceManager? .getInstance(configDir, configFile);
3. Record a meeting
4. Import the meeting into the CSLI KB (note that filterSausageData.pl is now executed automatically from the importer and calling it is no longer needed)
   • ImportCaloY3Recordings importer = new ImportCaloY3Recordings? ();
   • importer.run(new GenericProgressPanel? ());
5. Do action item processing
   • ActionItemIdentifierService aiService = csliServiceManager.getActionItemIdentifierService();
   • aiService.processOffline(meetingName);
6. Do topic processing
   • TopicSegmenterService tss = csliServiceManager.getTopicSegmenterService();
   • tss.prepareForNewMeeting(meetingName);
7. Use the browser to produce feedback
   • Clicking on the commit button will automatically cause retraining
8. After feedback has been committed, retrain the action item service
   • aiService.trainOffline();