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

A long standing goal of the data management community is to develop general,automated systems that ingest semi-structured documents and output queryabletables without human effort or domain specific customization. Given the sheervariety of potential documents, state-of-the art systems make simplifyingassumptions and use domain specific training. In this work, we ask whether wecan maintain generality by using large language models (LLMs). LLMs, which arepretrained on broad data, can perform diverse downstream tasks simplyconditioned on natural language task descriptions.  We propose and evaluate EVAPORATE, a simple, prototype system powered byLLMs. We identify two fundamentally different strategies for implementing thissystem: prompt the LLM to directly extract values from documents or prompt theLLM to synthesize code that performs the extraction. Our evaluations show acost-quality tradeoff between these two approaches. Code synthesis is cheap,but far less accurate than directly processing each document with the LLM. Toimprove quality while maintaining low cost, we propose an extended codesynthesis implementation, EVAPORATE-CODE+, which achieves better quality thandirect extraction. Our key insight is to generate many candidate functions andensemble their extractions using weak supervision. EVAPORATE-CODE+ not onlyoutperforms the state-of-the art systems, but does so using a sublinear passover the documents with the LLM. This equates to a 110x reduction in the numberof tokens the LLM needs to process, averaged across 16 real-world evaluationsettings of 10k documents each.

Language Models Enable Simple Systems for Generating Structured Views of Heterogeneous Data Lakes