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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

As one popular modeling approach for end-to-end speech recognition,attention-based encoder-decoder models are known to suffer the length bias andcorresponding beam problem. Different approaches have been applied in simplebeam search to ease the problem, most of which are heuristic-based and requireconsiderable tuning. We show that heuristics are not proper modelingrefinement, which results in severe performance degradation with largelyincreased beam sizes. We propose a novel beam search derived fromreinterpreting the sequence posterior with an explicit length modeling. Byapplying the reinterpreted probability together with beam pruning, the obtainedfinal probability leads to a robust model modification, which allows reliablecomparison among output sequences of different lengths. Experimentalverification on the LibriSpeech corpus shows that the proposed approach solvesthe length bias problem without heuristics or additional tuning effort. Itprovides robust decision making and consistently good performance under bothsmall and very large beam sizes. Compared with the best results of theheuristic baseline, the proposed approach achieves the same WER on the 'clean'sets and 4% relative improvement on the 'other' sets. We also show that it ismore efficient with the additional derived early stopping criterion.

Robust Beam Search for Encoder-Decoder Attention Based Speech Recognition without Length Bias