Search papers, labs, and topics across Lattice.
This study investigates the impact of conversational timing on the effectiveness of synthetic training data for automatic speech recognition (ASR) systems by treating timing properties as controllable variables. Utilizing an exponential-tilting family to parameterize pause and overlap distributions, the authors employ Latin hypercube sampling and multi-objective Bayesian optimization to explore the resulting parameter space. The findings reveal that ASR performance is significantly influenced by timing configurations, with increased overlap exposure leading to lower word error rates, underscoring the importance of timing diagnostics in training data generation.
Overlap in conversational training data can significantly reduce ASR error rates, revealing a critical trade-off between overlap and gap timing that reshapes data generation strategies.
Synthetic multi-speaker conversations are widely used to train conversational automatic speech recognition (ASR) systems, but it remains unclear which timing properties make simulated data most useful. This paper studies conversational timing as a controllable training variable rather than merely as a corpus statistic to be reproduced. We parameterize pause and overlap timing distributions with an exponential-tilting family estimated from multiple conversational corpora, and then explore the resulting four-dimensional parameter space with Latin hypercube sampling and multi-objective Bayesian optimization. Each sampled timing configuration is used to generate simulated training conversations, train an ASR system, and evaluate concatenated-permutation word and character error rates (cpWER and cpCER) on a Hungarian dialogue corpus. The results show that downstream ASR behavior is explained more directly by induced timing statistics than by raw simulator coordinates or corpus proximity. In particular, higher overlap exposure is associated with lower cpWER, whereas longer and more variable gaps are associated with higher cpWER; cpCER follows the same trend, but with weaker statistical support. Bayesian optimization yields modest aggregate improvements, but its main value is analytical: it produces controlled timing interventions that reveal an overlap--gap trade-off in simulated conversational training data. These findings suggest that realistic simulation should be complemented by task-relevant diagnostics of overlap, gap, and timing-variability profiles.
