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 Tools annual

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 monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Recent end-to-end automatic speech recognition (ASR) systems often utilize aTransformer-based acoustic encoder that generates embedding at a high framerate. However, this design is inefficient, particularly for long speech signalsdue to the quadratic computation of self-attention. To address this, we proposea new method, Adjacent Token Merging (A-ToMe), which gradually combinesadjacent tokens with high similarity scores between their key values. In thisway, the total time step could be reduced, and the inference of both theencoder and joint network is accelerated. Experiments on LibriSpeech show thatour method can reduce 57% of tokens and improve the inference speed on GPU by70% without any notable loss of accuracy. Additionally, we demonstrate thatA-ToMe is also an effective solution to reduce tokens in long-form ASR, wherethe input speech consists of multiple utterances.

Accelerating Transducers through Adjacent Token Merging