Feature analysis for discriminative confidence estimation in spoken term detection

Discriminative condence based on multi-layer perceptrons (MLPs) and multiple features has shown signicant advantage compared to the widely used lattice-based condence in spoken term detection (STD). Although the MLP-based framework can handle any features derived from a multitude of sources, choosing all possible features may lead to over complex models and hence less generality. In this paper, we design an extensive set of features and analyze their contribution to STD individually and as a group. The main goal is to choose a small set of features that are suciently informative while keeping the model simple and generalizable. We employ two established models to conduct the analysis: one is linear regression which targets for the most relevant features and the other is logistic linear regression which targets for the most discriminative features. We nd the most informative features are comprised of those derived from diverse sources (ASR decoding, duration and lexical properties) and the two models deliver highly consistent feature ranks. STD experiments on both English and Spanish data demonstrate signicant performance gains with the proposed feature sets.