Text Modular Networks: Learning to Decompose Tasks in the Language of Existing Models
This addresses the problem of interpretability and robustness in AI systems for multi-hop reasoning, offering a general approach that is more versatile and trustworthy than prior methods.
The paper tackles complex question answering by proposing Text Modular Networks (TMNs), a framework that learns to decompose tasks into simpler sub-questions solvable by existing models, resulting in a system (ModularQA) that shows improved versatility, robustness, and explanation quality on DROP and HotpotQA datasets.
We propose a general framework called Text Modular Networks(TMNs) for building interpretable systems that learn to solve complex tasks by decomposing them into simpler ones solvable by existing models. To ensure solvability of simpler tasks, TMNs learn the textual input-output behavior (i.e., language) of existing models through their datasets. This differs from prior decomposition-based approaches which, besides being designed specifically for each complex task, produce decompositions independent of existing sub-models. Specifically, we focus on Question Answering (QA) and show how to train a next-question generator to sequentially produce sub-questions targeting appropriate sub-models, without additional human annotation. These sub-questions and answers provide a faithful natural language explanation of the model's reasoning. We use this framework to build ModularQA, a system that can answer multi-hop reasoning questions by decomposing them into sub-questions answerable by a neural factoid single-span QA model and a symbolic calculator. Our experiments show that ModularQA is more versatile than existing explainable systems for DROP and HotpotQA datasets, is more robust than state-of-the-art blackbox (uninterpretable) systems, and generates more understandable and trustworthy explanations compared to prior work.