Assessing Language Models with Scaling Properties
This provides a new evaluation framework for language models, though it is incremental as it builds on existing scaling property concepts.
The authors tackled the problem of evaluating language models beyond perplexity by proposing tests based on natural language's scaling properties, finding that only neural language models exhibited long memory properties to a limited degree.
Language models have primarily been evaluated with perplexity. While perplexity quantifies the most comprehensible prediction performance, it does not provide qualitative information on the success or failure of models. Another approach for evaluating language models is thus proposed, using the scaling properties of natural language. Five such tests are considered, with the first two accounting for the vocabulary population and the other three for the long memory of natural language. The following models were evaluated with these tests: n-grams, probabilistic context-free grammar (PCFG), Simon and Pitman-Yor (PY) processes, hierarchical PY, and neural language models. Only the neural language models exhibit the long memory properties of natural language, but to a limited degree. The effectiveness of every test of these models is also discussed.