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

Passage retrieval is a fundamental task in many information systems, such asweb search and question answering, where both efficiency and effectiveness arecritical concerns. In recent years, neural retrievers based on pre-trainedlanguage models (PLM), such as dual-encoders, have achieved huge success. Yet,studies have found that the performance of dual-encoders are often limited dueto the neglecting of the interaction information between queries and candidatepassages. Therefore, various interaction paradigms have been proposed toimprove the performance of vanilla dual-encoders. Particularly, recentstate-of-the-art methods often introduce late-interaction during the modelinference process. However, such late-interaction based methods usually bringextensive computation and storage cost on large corpus. Despite theireffectiveness, the concern of efficiency and space footprint is still animportant factor that limits the application of interaction-based neuralretrieval models. To tackle this issue, we incorporate implicit interactioninto dual-encoders, and propose I^3 retriever. In particular, our implicitinteraction paradigm leverages generated pseudo-queries to simulatequery-passage interaction, which jointly optimizes with query and passageencoders in an end-to-end manner. It can be fully pre-computed and cached, andits inference process only involves simple dot product operation of the queryvector and passage vector, which makes it as efficient as the vanilla dualencoders. We conduct comprehensive experiments on MSMARCO and TREC2019 DeepLearning Datasets, demonstrating the I^3 retriever's superiority in terms ofboth effectiveness and efficiency. Moreover, the proposed implicit interactionis compatible with special pre-training and knowledge distillation for passageretrieval, which brings a new state-of-the-art performance.

I^3 Retriever: Incorporating Implicit Interaction in Pre-trained Language Models for Passage Retrieval