Enriching Under-Represented Named Entities for Improved Speech Recognition

Automatic speech recognition (ASR) for under-represented named-entity (UR-NE) is challenging due to such named-entities (NE) have insufficient instances and poor contextual coverage in the training data to learn reliable estimates and representations
^1
1
In this paper, UR-NE refers to the named-entity (NE) words that have low-frequency count, say, the count is in [1], [9] in this work, or do not appear in the training data at all, i.e. the count is 0.. In this paper, we propose approaches to enriching UR-NEs to improve speech recognition performance. Specifically, our first priority is to ensure those UR-NEs to appear in the word lattice if there is any. To this end, we employ class-based language model (LM) philosophy, making exemplar utterances for those UR-NEs according to their classes (e.g, location, person, organization, etc.), ending up with an improved LM that boosts the UR-NE occurrence in the word lattice. Then we boost the recognition performance through lattice rescoring methods. We first enrich the representations of UR-NEs in a pretrained recurrent neural network LM (RNNLM) by borrowing the embedding representations of the rich-represented NEs (RR-NEs), yielding the lattices that statistically favor the UR-NEs. Finally, we directly boost the likelihood scores of the utterances containing UR-NEs and gain further performance improvement.