Adversarial Bootstrapping for Dialogue Model Training

no code implementations • 3 Sep 2019 • Oluwatobi Olabiyi, Erik T. Mueller, Christopher Larson, Tarek Lahlou

Our experiments shows that adversarial bootstrapping is effective at addressing exposure bias, leading to improvement in response relevance and coherence.

Telephonetic: Making Neural Language Models Robust to ASR and Semantic Noise

no code implementations • 13 Jun 2019 • Chris Larson, Tarek Lahlou, Diana Mingels, Zachary Kulis, Erik Mueller

While techniques exist for desensitizing features to common noise patterns produced by Speech-to-Text (STT) and Text-to-Speech (TTS) systems, the question remains how to best leverage state-of-the-art language models (which capture rich semantic features, but are trained on only written text) on inputs with ASR errors.

Data Augmentation Decoder +1