Information on the 2018 Challenge

Overview

The 2018 FEIII challenge task is to enhance a given “ego” network dataset about publicly traded (S&P 500 index) financial entities, and their relationships to other entities. A simple enhancement could be to confirm a known relationship or to validate a predicted relationship, with respect to some ground truth dataset. A more valuable enhancement would be to predict a new and unknown relationship.

The ground truth dataset is currently a sample from the Thomson Reuters Data Fusion (TRDF) knowledge graph. Alternate ground truth datasets may be available from Bloomberg or IBM.

Datasets

We will pick a set of seed companies (financial entities) from NAICS sectors 52 (Finance and Insurance) and 51 (Information) to create the challenge dataset. Data that will be provided include the following:

Visualization of an exemplar dataset of 8 seed entities: https://karsha.umiacs.umd.edu/FinNetwork/

Task

The following “enhancement” tasks are listed in increasing order of difficulty / value of the enhancement:

Confirm a known relationship instance in the TRDF knowledge graph: The relationship already has 100% confidence in the TRDF graph.

Validate a predicted relationship instance in the TRDF knowledge graph: The relationship has < 100% confidence in the TR graph.

Enhance a known relationship instance in the TRDF knowledge graph: Add some additional properties to further describe the relationship.

Create a new relationship instance in the TRDF knowledge graph: The relationship is unknown in the TRDF graph.

Challenge participants may use additional external resources that include both properties and relationships for these financial entities. We note that an enhancement from a participant who does not use additional resources will be judged of greater value, in comparison to an enhancement that relied on external resources.

Tentative timeline

Friday March 16 Dataset and TRDF knowledge graph (training data) available for registered participants.
Friday March 30 Abstract submission to DSMM Workshop
Late April - Early May Release of TRDF test dataset and scoring of participant solutions.
Friday May 18 Short paper submission to DSMM Workshop
Friday June 15 DSMM Workshop