Compositional Attention Networks for Interpretability in Natural Language Question Answering
This work addresses interpretability in question answering for AI researchers, but it is incremental as it modifies an existing architecture for a new domain.
The paper tackles the problem of natural language question answering by adapting the MAC Net architecture from visual to natural language tasks, demonstrating its effectiveness on 20 bAbI tasks as a data-efficient and interpretable model.
MAC Net is a compositional attention network designed for Visual Question Answering. We propose a modified MAC net architecture for Natural Language Question Answering. Question Answering typically requires Language Understanding and multi-step Reasoning. MAC net's unique architecture - the separation between memory and control, facilitates data-driven iterative reasoning. This makes it an ideal candidate for solving tasks that involve logical reasoning. Our experiments with 20 bAbI tasks demonstrate the value of MAC net as a data-efficient and interpretable architecture for Natural Language Question Answering. The transparent nature of MAC net provides a highly granular view of the reasoning steps taken by the network in answering a query.