Abstract

Tree-structured recursive neural networks (TreeRNNs) for sentence meaning have been successful for many applications, but it remains an open question whether the fixed-length representations that they learn can support tasks as demanding as logi-cal deduction. We pursue this question by evaluating whether two such models— plain TreeRNNs and tree-structured neural tensor networks (TreeRNTNs)—can cor-rectly learn to identify logical relation-ships such as entailment and contradiction using these representations. In our first set of experiments, we generate artificial data from a logical grammar and use it to eval-uate the models ’ ability to learn to handle basic relational reasoning, recursive struc-tures, and quantification. We then evaluate the models on the more natural SICK chal-lenge data. Both models perform compet-itively on the SICK data and generalize well in all three experiments on simulated data, suggesting that they can learn suit-able representations for logical inference in natural language.