Transparency at the Source: Evaluating and Interpreting Language Models With Access to the True Distribution
This work provides a novel framework for researchers in natural language processing to rigorously assess model performance and interpretability, though it is incremental in its methodological approach.
The authors tackled the problem of evaluating and interpreting neural language models by using artificial, language-like data generated from a probabilistic grammar, which allowed them to compute exact lower bounds on perplexity and compare learned representations to symbolic rules. Their results revealed significant differences in how various architectures and training objectives approximate these bounds and in learning dynamics across word classes.
We present a setup for training, evaluating and interpreting neural language models, that uses artificial, language-like data. The data is generated using a massive probabilistic grammar (based on state-split PCFGs), that is itself derived from a large natural language corpus, but also provides us complete control over the generative process. We describe and release both grammar and corpus, and test for the naturalness of our generated data. This approach allows us to define closed-form expressions to efficiently compute exact lower bounds on obtainable perplexity using both causal and masked language modelling. Our results show striking differences between neural language modelling architectures and training objectives in how closely they allow approximating the lower bound on perplexity. Our approach also allows us to directly compare learned representations to symbolic rules in the underlying source. We experiment with various techniques for interpreting model behaviour and learning dynamics. With access to the underlying true source, our results show striking differences and outcomes in learning dynamics between different classes of words.