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/

Zendy Plus

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

Zendy Tools

Zendy Open

Industrial search and recommendation systems mostly follow the classicmulti-stage information retrieval paradigm: matching, pre-ranking, ranking, andre-ranking stages. To account for system efficiency, simple vector-productbased models are commonly deployed in the pre-ranking stage. Recent worksconsider distilling the high knowledge of large ranking models to smallpre-ranking models for better effectiveness. However, two major challenges inpre-ranking system still exist: (i) without explicitly modeling the performancegain versus computation cost, the predefined latency constraint in thepre-ranking stage inevitably leads to suboptimal solutions; (ii) transferringthe ranking teacher's knowledge to a pre-ranking student with a predeterminedhandcrafted architecture still suffers from the loss of model performance. Inthis work, a novel framework AutoFAS is proposed which jointly optimizes theefficiency and effectiveness of the pre-ranking model: (i) AutoFAS for thefirst time simultaneously selects the most valuable features and networkarchitectures using Neural Architecture Search (NAS) technique; (ii) equippedwith ranking model guided reward during NAS procedure, AutoFAS can select thebest pre-ranking architecture for a given ranking teacher without anycomputation overhead. Experimental results in our real world search system showAutoFAS consistently outperforms the previous state-of-the-art (SOTA)approaches at a lower computing cost. Notably, our model has been adopted inthe pre-ranking module in the search system of Meituan, bringing significantimprovements.

AutoFAS: Automatic Feature and Architecture Selection for Pre-Ranking System