Examination and Extension of Strategies for Improving Personalized Language Modeling via Interpolation (2006.05469v1)

Abstract: In this paper, we detail novel strategies for interpolating personalized LLMs and methods to handle out-of-vocabulary (OOV) tokens to improve personalized LLMs. Using publicly available data from Reddit, we demonstrate improvements in offline metrics at the user level by interpolating a global LSTM-based authoring model with a user-personalized n-gram model. By optimizing this approach with a back-off to uniform OOV penalty and the interpolation coefficient, we observe that over 80% of users receive a lift in perplexity, with an average of 5.2% in perplexity lift per user. In doing this research we extend previous work in building NLIs and improve the robustness of metrics for downstream tasks.