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

Transformer encoding networks have been proved to be a powerful tool ofunderstanding natural languages. They are playing a critical role in native adsservice, which facilitates the recommendation of appropriate ads based onuser's web browsing history. For the sake of efficient recommendation,conventional methods would generate user and advertisement embeddingsindependently with a siamese transformer encoder, such that approximate nearestneighbour search (ANN) can be leveraged. Given that the underlying semanticabout user and ad can be complicated, such independently generated embeddingsare prone to information loss, which leads to inferior recommendation quality.Although another encoding strategy, the cross encoder, can be much moreaccurate, it will lead to huge running cost and become infeasible for realtimeservices, like native ads recommendation. In this work, we propose hybridencoder, which makes efficient and precise native ads recommendation throughtwo consecutive steps: retrieval and ranking. In the retrieval step, user andad are encoded with a siamese component, which enables relevant candidates tobe retrieved via ANN search. In the ranking step, it further represents each adwith disentangled embeddings and each user with ad-related embeddings, whichcontributes to the fine-grained selection of high-quality ads from thecandidate set. Both steps are light-weighted, thanks to the pre-computed andcached intermedia results. To optimize the hybrid encoder's performance in thistwo-stage workflow, a progressive training pipeline is developed, which buildsup the model's capability in the retrieval and ranking task step-by-step. Thehybrid encoder's effectiveness is experimentally verified: with very littleadditional cost, it outperforms the siamese encoder significantly and achievescomparable recommendation quality as the cross encoder.

Hybrid Encoder: Towards Efficient and Precise Native AdsRecommendation via Hybrid Transformer Encoding Networks