Towards Automatic Topical Question Generation

We address the challenge of automatically generating questions from topics. We consider that each topic is associated with a body of texts containing useful information about the topic. Questions are generated by exploiting the named entity information and the predicate argument structures of the sentences present in the body of texts. To measure the importance of the generated questions, we use Latent Dirichlet Allocation (LDA) to identify the sub-topics (which are closely related to the original topic) in the given body of texts and apply the Extended String Subsequence Kernel (ESSK) to calculate their similarity with the questions. We also propose the use of syntactic tree kernels for computing the syntactic correctness of the questions. The questions are ranked by considering their importance (in the context of the given body of texts) and syntactic correctness. To the best of our knowledge, no other study has accomplished this task in our setting before. Experiments show that our approach can significantly outperform the state-of-the-art results.