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

Existing aspect extraction methods mostly rely on explicit or ground truthaspect information, or using data mining or machine learning approaches toextract aspects from implicit user feedback such as user reviews. It howeverremains under-explored how the extracted aspects can help generate moremeaningful recommendations to the users. Meanwhile, existing research onaspect-based recommendations often relies on separate aspect extraction modelsor assumes the aspects are given, without accounting for the fact the optimalset of aspects could be dependent on the recommendation task at hand.  In this work, we propose to combine aspect extraction together withaspect-based recommendations in an end-to-end manner, achieving the two goalstogether in a single framework. For the aspect extraction component, weleverage the recent advances in large language models and design a new promptlearning mechanism to generate aspects for the end recommendation task. For theaspect-based recommendation component, the extracted aspects are concatenatedwith the usual user and item features used by the recommendation model. Therecommendation task mediates the learning of the user embeddings and itemembeddings, which are used as soft prompts to generate aspects. Therefore, theextracted aspects are personalized and contextualized by the recommendationtask. We showcase the effectiveness of our proposed method through extensiveexperiments on three industrial datasets, where our proposed frameworksignificantly outperforms state-of-the-art baselines in both the personalizedaspect extraction and aspect-based recommendation tasks. In particular, wedemonstrate that it is necessary and beneficial to combine the learning ofaspect extraction and aspect-based recommendation together. We also conductextensive ablation studies to understand the contribution of each designcomponent in our framework.

Prompt Tuning Large Language Models on Personalized Aspect Extraction for Recommendations