Sequence-to-Sequence Networks Learn the Meaning of Reflexive Anaphora

TL;DR

This paper demonstrates that sequence-to-sequence recurrent networks can learn and generalize the meaning of reflexive anaphora in context, challenging previous doubts about their semantic capabilities.

Contribution

It shows that such networks can acquire semantic interpretations for reflexive anaphora and generalize to new antecedents, influenced by attention mechanisms and training data diversity.

Findings

Networks can generalize reflexive meanings to novel antecedents.

Attention mechanisms affect the learning process.

Training data diversity influences generalization success.

Abstract

Reflexive anaphora present a challenge for semantic interpretation: their meaning varies depending on context in a way that appears to require abstract variables. Past work has raised doubts about the ability of recurrent networks to meet this challenge. In this paper, we explore this question in the context of a fragment of English that incorporates the relevant sort of contextual variability. We consider sequence-to-sequence architectures with recurrent units and show that such networks are capable of learning semantic interpretations for reflexive anaphora which generalize to novel antecedents. We explore the effect of attention mechanisms and different recurrent unit types on the type of training data that is needed for success as measured in two ways: how much lexical support is needed to induce an abstract reflexive meaning (i.e., how many distinct reflexive antecedents must occur during training) and what contexts must a noun phrase occur in to support generalization of reflexive interpretation to this noun phrase?