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

Efficient document retrieval heavily relies on the technique of semantichashing, which learns a binary code for every document and employs Hammingdistance to evaluate document distances. However, existing semantic hashingmethods are mostly established on outdated TFIDF features, which obviously donot contain lots of important semantic information about documents.Furthermore, the Hamming distance can only be equal to one of several integervalues, significantly limiting its representational ability for documentdistances. To address these issues, in this paper, we propose to leverage BERTembeddings to perform efficient retrieval based on the product quantizationtechnique, which will assign for every document a real-valued codeword from thecodebook, instead of a binary code as in semantic hashing. Specifically, wefirst transform the original BERT embeddings via a learnable mapping and feedthe transformed embedding into a probabilistic product quantization module tooutput the assigned codeword. The refining and quantizing modules can beoptimized in an end-to-end manner by minimizing the probabilistic contrastiveloss. A mutual information maximization based method is further proposed toimprove the representativeness of codewords, so that documents can be quantizedmore accurately. Extensive experiments conducted on three benchmarksdemonstrate that our proposed method significantly outperforms currentstate-of-the-art baselines.

Efficient Document Retrieval by End-to-End Refining and Quantizing BERT Embedding with Contrastive Product Quantization