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

LSTM language models (LSTM-LMs) have been proven to be powerful and yieldedsignificant performance improvements over count based n-gram LMs in modernspeech recognition systems. Due to its infinite history states andcomputational load, most previous studies focus on applying LSTM-LMs in thesecond-pass for rescoring purpose. Recent work shows that it is feasible andcomputationally affordable to adopt the LSTM-LMs in the first-pass decodingwithin a dynamic (or tree based) decoder framework. In this work, the LSTM-LMis composed with a WFST decoder on-the-fly for the first-pass decoding.Furthermore, motivated by the long-term history nature of LSTM-LMs, the use ofcontext beyond the current utterance is explored for the first-pass decoding inconversational speech recognition. The context information is captured by thehidden states of LSTM-LMs across utterance and can be used to guide thefirst-pass search effectively. The experimental results in our internal meetingtranscription system show that significant performance improvements can beobtained by incorporating the contextual information with LSTM-LMs in thefirst-pass decoding, compared to applying the contextual information in thesecond-pass rescoring.

LSTM-LM with Long-Term History for First-Pass Decoding in Conversational Speech Recognition