Language Model Evaluation Beyond Perplexity
This work addresses the need for more nuanced evaluation metrics in natural language processing, though it is incremental as it builds on existing statistical analysis methods.
The authors tackled the problem of evaluating language models beyond perplexity by assessing how well generated text matches statistical tendencies of natural language, finding that models learn only a subset of these tendencies and alignment varies with architecture and generation strategies, such as nucleus sampling outperforming ancestral sampling in type-token relationships.
We propose an alternate approach to quantifying how well language models learn natural language: we ask how well they match the statistical tendencies of natural language. To answer this question, we analyze whether text generated from language models exhibits the statistical tendencies present in the human-generated text on which they were trained. We provide a framework--paired with significance tests--for evaluating the fit of language models to these trends. We find that neural language models appear to learn only a subset of the tendencies considered, but align much more closely with empirical trends than proposed theoretical distributions (when present). Further, the fit to different distributions is highly-dependent on both model architecture and generation strategy. As concrete examples, text generated under the nucleus sampling scheme adheres more closely to the type--token relationship of natural language than text produced using standard ancestral sampling; text from LSTMs reflects the natural language distributions over length, stopwords, and symbols surprisingly well.