A crowdsourced Who wants to be a millionaire? player

Question answering is a fundamental problem for artificial intelligence research. State-of-the-art question answering systems, such as search engines, can answer well-formed factual questions. However, they fail on nonfactual and natural language queries. On the other hand, crowdsourcing leverages human intelligence to present solutions for problems, which are hard for computers. In the process of crowdsourcing, aggregating the responses from the crowd is a big challenge of itself. This work presents a model that integrates artificial intelligence with crowdsourcing to answer difficult natural language multiple choice questions. We build a crowdsourced mobile gaming experience for Who wants to be a millionaire? TV quiz show to test our methods. We use lightweight artificial intelligence models in aggregating the crowd responses. Our methods are able to answer even the hardest questions with very high accuracy. The experiment results suggest that building a super player for Who wants to be a millionaire? is feasible using our models. This study shows that facilitating crowdsourcing with artificial intelligence is an important tool for answering questions, which trouble state-of-the-art question answering systems.

Automated summary

This study combines artificial intelligence and crowdsourcing to answer difficult natural language multiple choice questions, successfully demonstrating that even the hardest questions can be answered using this approach.