Deep Graph Random Process for Relational-Thinking-Based Speech Recognition
This addresses the challenge of incorporating relational thinking into ASR for improved conversational understanding, representing an incremental advancement by applying a novel method to a known bottleneck in the field.
The paper tackled the problem of modeling relational thinking for conversational automatic speech recognition by proposing a Bayesian nonparametric deep learning method that generates infinite probabilistic graphs representing percepts and provides a closed-form solution for coupling them, achieving effectiveness in ASR tasks on CHiME-2 and CHiME-5 datasets.
Lying at the core of human intelligence, relational thinking is characterized by initially relying on innumerable unconscious percepts pertaining to relations between new sensory signals and prior knowledge, consequently becoming a recognizable concept or object through coupling and transformation of these percepts. Such mental processes are difficult to model in real-world problems such as in conversational automatic speech recognition (ASR), as the percepts (if they are modelled as graphs indicating relationships among utterances) are supposed to be innumerable and not directly observable. In this paper, we present a Bayesian nonparametric deep learning method called deep graph random process (DGP) that can generate an infinite number of probabilistic graphs representing percepts. We further provide a closed-form solution for coupling and transformation of these percept graphs for acoustic modeling. Our approach is able to successfully infer relations among utterances without using any relational data during training. Experimental evaluations on ASR tasks including CHiME-2 and CHiME-5 demonstrate the effectiveness and benefits of our method.