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

An advantage of seq2seq abstractive summarization models is that theygenerate text in a free-form manner, but this flexibility makes it difficult tointerpret model behavior. In this work, we analyze summarization decoders inboth blackbox and whitebox ways by studying on the entropy, or uncertainty, ofthe model's token-level predictions. For two strong pre-trained models, PEGASUSand BART on two summarization datasets, we find a strong correlation betweenlow prediction entropy and where the model copies tokens rather than generatingnovel text. The decoder's uncertainty also connects to factors like sentenceposition and syntactic distance between adjacent pairs of tokens, giving asense of what factors make a context particularly selective for the model'snext output token. Finally, we study the relationship of decoder uncertaintyand attention behavior to understand how attention gives rise to these observedeffects in the model. We show that uncertainty is a useful perspective foranalyzing summarization and text generation models more broadly.

Understanding Neural Abstractive Summarization Models via Uncertainty