English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

This paper presents our modeling and architecture approaches for building ahighly accurate low-latency language identification system to supportmultilingual spoken queries for voice assistants. A common approach to solvemultilingual speech recognition is to run multiple monolingual ASR systems inparallel and rely on a language identification (LID) component that detects theinput language. Conventionally, LID relies on acoustic only information todetect input language. We propose an approach that learns and combines acousticlevel representations with embeddings estimated on ASR hypotheses resulting inup to 50% relative reduction of identification error rate, compared to a modelthat uses acoustic only features. Furthermore, to reduce the processing costand latency, we exploit a streaming architecture to identify the spokenlanguage early when the system reaches a predetermined confidence level,alleviating the need to run multiple ASR systems until the end of input query.The combined acoustic and text LID, coupled with our proposed streaming runtimearchitecture, results in an average of 1500ms early identification for morethan 50% of utterances, with almost no degradation in accuracy. We also showimproved results by adopting a semi-supervised learning (SSL) technique usingthe newly proposed model architecture as a teacher model.

Streaming Language Identification using Combination of Acoustic Representations and ASR Hypotheses