Background

       To promote academic and practical research on big data, Chinese Association for Artificial Intelligence, IEEE-China and Toutiao are organizing an International Machine Learning Competition. The challenge problem is to match community raised questions with domain experts. Toutiao’s platform has accumulated 580 million users and 300 thousand professional freelance writers. Toutiao has been extensively developing large-scale machine learning systems for personalized news, picture and video recommendation.
       Toutiao Q&A is an up-and-coming mobile social platform, built upon 530 million Toutiao users and precise recommendation algorithm, promoting short-form content creation and interaction on mobile devices in the format of Q&A. We strive to match information with the right people, finding the best respondent to the questions, and the best readers to the answers.
       Toutiao Q&A now has tens of thousand users answering on a daily basis, while user-created answers are viewed tens of millions times a day. An important question is how to match questions with interested users’ expertise. If the matching strategy is not accurate enough, to ensure that questions get enough top quality answers, the system will have to send out invitations to more expert users who may be disturbed.

Task

       Participating teams will use data provided by Toutiao Q&A to conduct analysis on specific questions. Each data record includes expert tags, question data and question distribution data. Given certain questions, participants need to forecast which experts are more likely to answer which questions. Specifically, given each question and each expert, participants need to calculate the probability of that expert answering the question.
       In the operation, the management system will invite experts to answer the question in the sequence of the probability, until a certain number of answers are obtained. In terms of evaluation, given a certain question, we will rank the experts based on the forecasted probability, and evaluate the NDCG@5 and NDCG@10 of ranking results. The final evaluation formula is:

       NDCG@5 * 0.5 + NDCG@10 * 0.5.

If have any questions about the datas or the task,please link to the FAQ page or send a message to bytecup2016@bytedance.com and wait for a reply