Each Contestant will have access to two sets of call data from an existing robocall honeypot. Contestants will be authorized to access the data only after the Contestant submits a Registration Form to the Sponsor. To access this data, Contestants may also be required to agree to certain terms and conditions specified by Pindrop. Contestants may not disseminate this data to any individual or legal entity other than another team member, and Contestants shall be disqualified from the Contest by the Sponsor if they do so.
The first data set will identify calls that are likely to have been a robocall (a call delivering a prerecorded message). Based on information provided in the first data set, Contestants will develop an algorithm and will predict which of the calls in the second data set are likely to be robocalls. The Contestants’ predictions must be submitted in ascii text format using CSV. In addition to submitting these predictions, each Contestant must submit all source code and a written description of the algorithm consisting of fewer than 250 words.
Judging Criteria
Uncovering the Truth - 70% of total score
How well did you predict whether the calls in the second honeypot data set were likely to be robocalls? To assess this, the Contest Judges will compare your predictions with actual data about which calls are likely to be a robocall in the second data set. You will receive one point for each call you successfully identified as a likely robocall, and deducted two points for each call you inaccurately identified as a likely robocall.
Explaining the Scheme - 20% of total score
What insights did your Submission demonstrate with respect to the analysis of honeypot call records?
Innovation - 10% of total score
How innovative was your Submission?
Complete judging criteria and other requirements are in the official Rules.