Few-Shot Complex Knowledge Base Question Answering via Meta Reinforcement Learning
This work addresses distributional bias in question types for complex QA, offering a few-shot learning solution that could improve efficiency in knowledge base systems, though it is incremental as it builds on existing neural program induction methods.
The paper tackles uneven performance in complex knowledge base question answering by proposing a meta-reinforcement learning approach that adapts to new questions using retrieved similar ones, achieving state-of-the-art results on the CQA dataset with only five trial trajectories and 1% of the training data for meta-training.
Complex question-answering (CQA) involves answering complex natural-language questions on a knowledge base (KB). However, the conventional neural program induction (NPI) approach exhibits uneven performance when the questions have different types, harboring inherently different characteristics, e.g., difficulty level. This paper proposes a meta-reinforcement learning approach to program induction in CQA to tackle the potential distributional bias in questions. Our method quickly and effectively adapts the meta-learned programmer to new questions based on the most similar questions retrieved from the training data. The meta-learned policy is then used to learn a good programming policy, utilizing the trial trajectories and their rewards for similar questions in the support set. Our method achieves state-of-the-art performance on the CQA dataset (Saha et al., 2018) while using only five trial trajectories for the top-5 retrieved questions in each support set, and metatraining on tasks constructed from only 1% of the training set. We have released our code at https://github.com/DevinJake/MRL-CQA.