Do as I mean, not as I say: Sequence Loss Training for Spoken Language Understanding
This work addresses the misalignment between training losses and evaluation metrics in SLU systems, which is crucial for improving accuracy in voice-activated applications, though it is incremental as it builds on existing reinforcement learning techniques.
The paper tackled the problem of training spoken language understanding (SLU) systems, which often use cross-entropy losses that do not align with performance metrics like word or semantic error rates, by proposing non-differentiable sequence losses based on SLU metrics and using REINFORCE for training. The result was a state-of-the-art method on open datasets and a 6% relative improvement in ASR and NLU metrics on large proprietary datasets, with the ability to update models using only semantic feedback.
Spoken language understanding (SLU) systems extract transcriptions, as well as semantics of intent or named entities from speech, and are essential components of voice activated systems. SLU models, which either directly extract semantics from audio or are composed of pipelined automatic speech recognition (ASR) and natural language understanding (NLU) models, are typically trained via differentiable cross-entropy losses, even when the relevant performance metrics of interest are word or semantic error rates. In this work, we propose non-differentiable sequence losses based on SLU metrics as a proxy for semantic error and use the REINFORCE trick to train ASR and SLU models with this loss. We show that custom sequence loss training is the state-of-the-art on open SLU datasets and leads to 6% relative improvement in both ASR and NLU performance metrics on large proprietary datasets. We also demonstrate how the semantic sequence loss training paradigm can be used to update ASR and SLU models without transcripts, using semantic feedback alone.