Welcome to Yu Yu's homepage!

We propose a neural language modeling system based on low-rank adaptation (LoRA) for speech recognition output rescoring. 

Conventional data selection methods select training samples based on the test domain knowledge and not on real life data, thus they frequently fail in unknown domains like patent and Twitter. We propose to select training samples that maximize information uncertainty measured by observation entropy like empirical Shannon entropy, and prediction entropy using mutual information, to cover more possible queries that may appear in unknown worlds.

This paper introduces our Diversity Advanced Actor-Critic reinforcement learning (A2C) framework (DAAC) to improve the generalization and accuracy of Natural Language Processing (NLP). We quantify diversity on a set of samples using the max dispersion, convex hull volume, and graph entropy based on sentence embeddings in high dimensional metric space. 

Unlike previous work that has defined robustness using Minimax to bound worst cases, we measure robustness based on the consistency of cross-domain accuracy and introduce the coefficient of variation and (ε, γ)-Robustness.