Under the Hood: Using Diagnostic Classifiers to Investigate and Improve how Language Models Track Agreement Information
This provides insights into linguistic representation in neural models, potentially aiding interpretability and error correction, though it is incremental in applying existing diagnostic methods to a specific linguistic task.
The paper tackled the problem of understanding how neural language models track subject-verb number agreement by using diagnostic classifiers to analyze internal states, and demonstrated that this knowledge can be used to improve model accuracy, with interventions leading to a large increase in performance.
How do neural language models keep track of number agreement between subject and verb? We show that `diagnostic classifiers', trained to predict number from the internal states of a language model, provide a detailed understanding of how, when, and where this information is represented. Moreover, they give us insight into when and where number information is corrupted in cases where the language model ends up making agreement errors. To demonstrate the causal role played by the representations we find, we then use agreement information to influence the course of the LSTM during the processing of difficult sentences. Results from such an intervention reveal a large increase in the language model's accuracy. Together, these results show that diagnostic classifiers give us an unrivalled detailed look into the representation of linguistic information in neural models, and demonstrate that this knowledge can be used to improve their performance.