Neural language models (NLM) have been shown to outperform conventional n-gram language models by a substantial margin in Automatic Speech Recognition (ASR) and other tasks. There are, however, a number of challenges that need to be addressed for an NLM to be used in a practical large-scale ASR system. In this paper, we present solutions to some of the challenges, including training NLM from heterogenous corpora, limiting latency impact and handling personalized bias in the second-pass rescorer. Overall, we show that we can achieve a 6.2% relative WER reduction using neural LM in a secondpass n-best rescoring framework with a minimal increase in latency.

No abstract

No abstract

We present a scalable joint language model designed to utilize fine-grain syntactic tags. We discuss challenges such a design faces and describe our solutions that scale well to large tagsets and corpora. We advocate the use of relatively simple tags that do not require deep linguistic knowledge of the language but provide more structural information than POS tags and can be derived from automatically generated parse trees-a combination of properties that allows easy adoption of this model for new languages. We propose two fine-grain tagsets and evaluate our model using these tags, as well as POS tags and SuperARV tags in a speech recognition task and discuss future directions.

The recognition of personalized content, such as contact names, remains a challenging problem for end-to-end speech recognition systems. In this work, we demonstrate how first-and second-pass rescoring strategies can be leveraged together to improve the recognition of such words. Following previous work, we use a shallow fusion approach to bias towards recognition of personalized content in the first-pass decoding. We show that such an approach can improve personalized content recognition by up to 16% with minimum degradation on the general use case. We describe a fast and scalable algorithm that enables our biasing models to remain at the wordlevel, while applying the biasing at the subword level. This has the advantage of not requiring the biasing models to be dependent on any subword symbol table. We also describe a novel second-pass de-biasing approach: used in conjunction with a first-pass shallow fusion that optimizes on oracle WER, we can achieve an additional 14% improvement on personalized content recognition, and even improve accuracy for the general use case by up to 2.5%.