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

Continual learning (CL), or domain expansion, recently became a popular topicfor automatic speech recognition (ASR) acoustic modeling because practicalsystems have to be updated frequently in order to work robustly on types ofspeech not observed during initial training. While sequential adaptation allowstuning a system to a new domain, it may result in performance degradation onthe old domains due to catastrophic forgetting. In this work we exploreregularization-based CL for neural network acoustic models trained with thelattice-free maximum mutual information (LF-MMI) criterion. We simulate domainexpansion by incrementally adapting the acoustic model on different publicdatasets that include several accents and speaking styles. We investigate twowell-known CL techniques, elastic weight consolidation (EWC) and learningwithout forgetting (LWF), which aim to reduce forgetting by preserving modelweights or network outputs. We additionally introduce a sequence-level LWFregularization, which exploits posteriors from the denominator graph of LF-MMIto further reduce forgetting. Empirical results show that the proposedsequence-level LWF can improve the best average word error rate across alldomains by up to 9.4% relative compared with using regular LWF.

Continual learning using lattice-free MMI for speech recognition