Composition-based on-the-fly rescoring for salient n-gram biasing

76 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Personalization of End-to-End Speech Recognition on Mobile Devices for Named Entities (2019, methodology): “This includes techniques such as biasing [7, 8, 9], shallow fusion [10], deep fusion [10] and cold fusion [11].”
Streaming End-to-end Speech Recognition for Mobile Devices (2018, methodology): “Conventional ASR systems perform contextual biasing by building an n-gram finite state transducer (FST) from a list of biasing phrases, which is composed on-the-fly with the decoder graph during decoding [36].”
Mobile Keyboard Input Decoding with Finite-State Transducers (2017, methodology): “Instead, they are incorporated in the decoding process using an on-the-fly lattice rescoring technique, similar to the n-gram biasing approach described in [13].”
Hybrid Autoregressive Transducer (HAT) (2020, methodology): “Another advantage of the modular modeling approach is it allows dynamic modification of the component models in well proven, principled methods such as vocabulary augmentation [5], LM adaptation and contextual biasing [6].”
Personalized speech recognition on mobile devices (2016, methodology): “In either we can perform the additional step of using on-the-ﬂy rescoring as in [13] to bias the language model towards recognizing only these contact names.”