Word-order biases in deep-agent emergent communication
This work addresses the problem of aligning neural network language processing with human linguistic biases for researchers in NLP and cognitive science, though it is incremental in exploring specific biases.
The study investigated whether sequence-processing neural networks exhibit human-like word-order biases, such as avoiding redundancy and minimizing long-distance dependencies, by training them to communicate in a gridworld with controlled miniature languages. Results showed a strong tendency to avoid long-distance dependencies but no clear preference for non-redundant encoding, suggesting the need to incorporate 'effort' into models for more human-like behavior.
Sequence-processing neural networks led to remarkable progress on many NLP tasks. As a consequence, there has been increasing interest in understanding to what extent they process language as humans do. We aim here to uncover which biases such models display with respect to "natural" word-order constraints. We train models to communicate about paths in a simple gridworld, using miniature languages that reflect or violate various natural language trends, such as the tendency to avoid redundancy or to minimize long-distance dependencies. We study how the controlled characteristics of our miniature languages affect individual learning and their stability across multiple network generations. The results draw a mixed picture. On the one hand, neural networks show a strong tendency to avoid long-distance dependencies. On the other hand, there is no clear preference for the efficient, non-redundant encoding of information that is widely attested in natural language. We thus suggest inoculating a notion of "effort" into neural networks, as a possible way to make their linguistic behavior more human-like.