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

Most neural machine translation systems are built upon subword unitsextracted by methods such as Byte-Pair Encoding (BPE) or wordpiece. However,the choice of number of merge operations is generally made by followingexisting recipes. In this paper, we conduct a systematic exploration ondifferent numbers of BPE merge operations to understand how it interacts withthe model architecture, the strategy to build vocabularies and the languagepair. Our exploration could provide guidance for selecting proper BPEconfigurations in the future. Most prominently: we show that for LSTM-basedarchitectures, it is necessary to experiment with a wide range of different BPEoperations as there is no typical optimal BPE configuration, whereas forTransformer architectures, smaller BPE size tends to be a typically optimalchoice. We urge the community to make prudent choices with subword mergeoperations, as our experiments indicate that a sub-optimal BPE configurationalone could easily reduce the system performance by 3-4 BLEU points.

A Call for Prudent Choice of Subword Merge Operations in Neural Machine Translation