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

In this work, we introduce back-training, an alternative to self-training forunsupervised domain adaptation (UDA) from source to target domain. Whileself-training generates synthetic training data where natural inputs arealigned with noisy outputs, back-training results in natural outputs alignedwith noisy inputs. This significantly reduces the gap between the target domainand synthetic data distribution, and reduces model overfitting to the sourcedomain. We run UDA experiments on question generation and passage retrievalfrom the \textit{Natural Questions} domain to machine learning and biomedicaldomains. We find that back-training vastly outperforms self-training by a meanimprovement of 7.8 BLEU-4 points on generation, and 17.6\% top-20 retrievalaccuracy across both domains. We further propose consistency filters to removelow-quality synthetic data before training. We also release a newdomain-adaptation dataset- \textit{MLQuestions} containing 35K unalignedquestions, 50K unaligned passages, and 3K aligned question-passage pairs.

Back-Training excels Self-Training at Unsupervised Domain Adaptation of Question Generation and Passage Retrieval