On the Challenges of Evaluating Compositional Explanations in Multi-Hop Inference: Relevance, Completeness, and Expert Ratings

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 7529–7542 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 7529 On the Challenges of Evaluating Compositional Explanations in Multi-Hop Inference: Relevance, Completeness, and Expert Ratings Peter A. Jansen and Kelly Smith and Dan Moreno and Huitzilin Ortiz University of Arizona, USA pajansen@arizona.edu Abstract Building compositional explanations requires models to combine two or more facts that, to- gether, describe why the answer to a question is correct. Typically, these “multi-hop” expla- nations are evaluated relative to one (or a small number of) gold explanations. In this work, we show these evaluations substantially underesti- mate model performance, both in terms of the relevance of included facts, as well as the com- pleteness of model-generated explanations, be- cause models regularly discover and produce valid explanations that are different than gold explanations. To address this, we construct a large corpus of 126k domain-expert (science teacher) relevance ratings that augment a cor- pus of explanations to standardized science exam questions, discovering 80k additional relevant facts not rated as gold. We build three strong models based on different methodolo- gies (generation, ranking, and schemas), and empirically show that while expert-augmented ratings provide better estimates of explana- tion quality, both original (gold) and expert- augmented automatic evaluations still substan- tially underestimate performance by up to 36% when compared with full manual expert judge- ments, with different models being dispropor- tionately affected. This poses a significant methodological challenge to accurately evalu- ating explanations produced by compositional reasoning models. 1 Introduction Compositional inference is the high-level task of combining two or more pieces of knowledge to perform reasoning. In the context of question answering, compositional (or “multi-hop”) infer- ence typically takes the form of combining facts from a knowledge base that allow a given solver to form a complete chain-of-reasoning that moves from question to correct answer. A desirable con- sequence is that the facts used to assemble this chain-of-reasoning can then be taken as an inter- Question: When trees are cleared from the land, what will most likely occur? Answer: Soil Erosion Gold Explanation A tree is a kind of plant. Roots are a part of a plant. In the soil erosion process, plant roots are an inhibitor. Removing an inhibitor causes that process to happen. Soil erosion is when wind/water move soil. Tree roots decrease soil erosion. As deforestation increases, soil erosion will increase. Deforested area is where humans cut down trees. Clearing a forest means cutting down trees. Model-Generated Explanation Automated Evaluation Expert Evaluation Figure 1: An example science exam question, its gold expla- nation from the WorldTree corpus, and a model-generated ex- planation from one of the models (Tensorflow-Ranking-BERT) trained using expert-generated relevance ratings produced in this work. Though the model-generated explanation is strong, it shares no facts in common with the gold explanation, and automatic evaluations rate it neither relevant nor complete. pretable record of that reasoning, as well as a human-readable explanation for why the answer is correct. Compositional inference …

Content truncated to 500 words