Skip to content

trecrts/trecrts-eval

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

65 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

TREC Real-Time Summarization Tools

Herein we detail the various tools to be used in the course of the TREC 2016 Real-Time Summarization track.

The track coordinators envision the interaction model outlined in rts-arch/trecrtstwitter.pdf

TREC participants write clients that interact with a broker run by the coordinators. This broker manages all communication with mobile assessors (i.e., humans with the mobile assessing application on their phone). The assessments rendered by the participants are then made available to the clients through a pull request.

Note that for the mobile assessment application and the broker, participants should not run their own!

Details for the Broker API can be found here[https://github.com/aroegies/trecrts-tools/tree/master/trecrts-server].

Broker Server (NOTE: Participants do not run this -- Track Coordinators do)

Requirements

  • Node.js v5.0.0+
  • npm v3.3.6+
  • MySQL Server v5.5+

Installation

  • Standard Node.js install procedure, run npm install from trecrts-server
    • You will need to supply a push_auths.js in trecrts-server/routes which exports the relevant gcm server details from the Google Developer console
  • MySQL setup is relatively straightforward
    • Create a trec_rts database
    • Load the template from trecrts-server/rts.db.template into the DB
    • Add appropriate group identifiers to the groups table
    • Add any new topics to the topics table and creating the corresponding judgements table and seen table for the topic based upon the judgements_template table and seen_template respectively
      • For example, create table judgements_MBFoo like judgements_template; and create table seen_MBFoo like seen_template;

Running the Broker

  • Standard Express.js invocation, npm start
    • As usual, you may supply a PORT argument beforehand to change the connecting port (e.g., PORT=10101 npm start)

Notes

  • The broker comes supplied with a web app, located at hostname:port/index.html, that connects to the broker to simulate the mobile assessor interface in the web browser. This may be more convenient for testing clients than deploying to phones.
    • Web app assessments are functionally annonymous

Mobile Assessment App (NOTE: Participants do not run this -- Track Coordinators do)

Requirements

Cordova v5.2+ Android device (currently, iOS soon)

Installation

Standard Cordova installation procedure:

Notes

  • If you want to use the mobile app with your own server you will need to edit the index.js file to point to the appropriate GCM service and the broker API host/port.

Participant Clients

trecrts-clients/ contains several simple client implementations to provide simple examples of how a RTS system might work. These examples have been structured to facilitate plug-and-play and so should not require many dependencies (e.g., complex machine learning packages).

A more meaningful example can be found in the Anserini[https://github.com/lintool/Anserini] project.