Learning to Fingerprint the Latent Structure in Question Articulation

Algorithmic interpretation of question articulation is one of the key steps in a machine-driven question-answering system. Machine understanding of an input question is tightly related to recognition of articulation in the context of the computational capabilities of an underlying processing algorithm. In this paper, a mathematical model to capture and distinguish the latent structure in the articulation of questions is presented. It is argued that an objective-driven approach to represent this latent structure is beneficial when examples of complementary objectives are not available. We show that the latent structure can be represented as a system that maximizes a cost function related to the underlying objective. Further, we show that the optimization formulation can be approximated to building a memory of patterns represented as a trained neural auto-encoder. Experimental evaluation using many clusters of questions, each related to an objective, shows 80% recognition accuracy and a negligible false positive across these clusters of questions. We then extend the same memory to a related task where the goal is to iteratively refine clusters of questions based on latent articulation. We show that the refinement scheme called K-fingerprints achieves nearly 100% cluster recognition with negligible false positive between different clusters.