Spoken Term Detection and Relevance Score Estimation using Dot-Product of Pronunciation Embeddings
This work addresses spoken term detection for large archives, representing an incremental improvement over prior Siamese network approaches.
The paper tackles spoken term detection in large spoken archives by extending Siamese neural networks with deep LSTM networks to project phoneme confusion networks and searched terms into an embedding space, computing relevance scores via dot-product and sigmoid calibration, and localizing terms; it achieves experimental evaluation on MALACH data in English and Czech.
The paper describes a novel approach to Spoken Term Detection (STD) in large spoken archives using deep LSTM networks. The work is based on the previous approach of using Siamese neural networks for STD and naturally extends it to directly localize a spoken term and estimate its relevance score. The phoneme confusion network generated by a phoneme recognizer is processed by the deep LSTM network which projects each segment of the confusion network into an embedding space. The searched term is projected into the same embedding space using another deep LSTM network. The relevance score is then computed using a simple dot-product in the embedding space and calibrated using a sigmoid function to predict the probability of occurrence. The location of the searched term is then estimated from the sequence of output probabilities. The deep LSTM networks are trained in a self-supervised manner from paired recognition hypotheses on word and phoneme levels. The method is experimentally evaluated on MALACH data in English and Czech languages.