This article investigates whether unsupervised dependency parsing can be evaluated without a gold standard in non-human species, where such reference data is typically unavailable.

  • The authors apply network science to demonstrate that the proportion of correct edges retrieved by a parser must be high for non-human primate sequences due to the fast decay of sequence length distribution.
  • In contrast, human language sequences lack this property, making evaluation without a gold standard a hard problem in humans but feasible in non-human primates.

The findings suggest that evaluation is possible for non-human primates despite the absence of gold standards, challenging the assumption that dependency parsing is unfeasible in other species.