Competition

Competition takes place on our online platfrom, specifically developed for Sberbank Data Science Journey. Participation rules are availabe here.

Problems

We prepared 2 problems of different difficulty level. Participants will receive a dataset which consists of text paragraphs, questions and their corresponding answers tuples. To download data and send your sudmissions please use our online platform.

A. Classify relevant questions

Binary classification. Given text paragraph and a set of questions, can you identify which question is real and was asked by a human person?

Difficulty: Medium
Evaluation metric: ROC-AUC
Submission format: Offline test-set labeling

B. Question answering system

Dialogue system. Can you build a question answering system that infers an exact text answer, given a paragraph of text and a relevant question?

Difficulty: High
Evaluation metric: (Macro-averaged) F1-score
Submission format: Docker container with inference model

Prizes

Total prize fund is 2 000 000* rubles. Winners are determined automatically.

Problem A

  • 1 place: 200 000* rubles
  • 2 place: 100 000* rubles
  • 3 place: 50 000* rubles

Winners are determined automatically based on their leaderboard positions on problem A and are publically announced right after the competition ends, November 1.

Problem B

  • 1 place: 1 000 000* rubles
  • 2 place: 200 000* rubles
  • 3 place: 100 000* rubles
  • 4 place: 50 000* rubles
  • 5 place: 50 000* rubles

Top-10 winners are determined automatically based on their leaderboard positions on problem B, but their exact order is announced on November 11 during Data Science Day.

Jury Prize

  • 1 place: 250 000* rubles

Top-10 winners on problem B are invited to take part in an interactive competition finale on November 11 during Data Science Day. During the finale Jury of invited top-managers of Sberbank and renowned experts will be asking questions to their dialgue systems and evaluate their quality. The system with highest Jury score will receive this extra 250 000* rubles prize.

Winner awards ceremony will take place on November 11 during Data Science Day.

*Personal Income Tax will be withheld, Sberbank acts as a tax agent.

FAQ

List of frequently asked questions is updated regularly, stay tuned!

How can I participate in Data Science Contest?

You have to register by filling out the form. Develop you problem solutions on the provided datasets. Submit your solution and see how it ranks among others. Solutions can be submitted again.

What are the solution evaluation criteria?

Problem solutions are graded automatically based on comparison to the true data labels in test data, available only for organizers. Leaderboard is evaluated online and updated in real time.

Is participation free?

Yes. Registration and participation are free.

Where will Data Science Contest be hosted?

Sberbank Data Science Contest will be hosted online on this website.

Who is allowed to participate in Data Science Contest?

Any person who is at least 18 years old is allowed to participate in competition.

Can people from other cities and countries participate?

Yes they can.
 

Are there any limitations on participating in Data Science Contest?

Members of the competition organization and dataset collection team as well as others directly or indirectly involved are not allowed to participate.

Are Sberbank company group employees allowed to participate in this competition?

Sberbank group employees are allowed to participate, but are not eligible for monetary prizes. They will have "[ooc]" out of competition tags in the leaderboard.

Will there be any award ceremony?

Yes. Winner awards ceremony and the competition finale will take place on November 11 during Data Science Day.

What are the prizes?

The list of prizes for each problem and the Jury prize awarded during competetion finale are listed here.

When will the registration open?

Registration will be opne from September 14 untill October 30 inclusive.

Can I join the competition later?

Yes. Solution submissions will be open until October 31 03:00:00 (GMT+3, Moscow time) inclusive.