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

Modern neural language models can produce remarkably fluent and grammaticaltext. So much, in fact, that recent work by Clark et al. (2021) has reportedthat conventional crowdsourcing can no longer reliably distinguish betweenmachine-authored (GPT-3) and human-authored writing. As errors in machinegenerations become ever subtler and harder to spot, it poses a new challenge tothe research community for robust machine text evaluation. We propose a newframework called Scarecrow for scrutinizing machine text via crowd annotation.To support the broad range of real machine errors that can be identified bylaypeople, the ten error categories of Scarecrow -- such as redundancy,commonsense errors, and incoherence -- are identified through several rounds ofcrowd annotation experiments without a predefined ontology. We then useScarecrow to collect over 41k error spans in human-written andmachine-generated paragraphs of English language news text. We isolate factorsfor detailed analysis, including parameter count, training data, and variousdecoding-time configurations. Our approach successfully quantifies measurablegaps between human authored text and generations from models of several sizes,including fourteen configurations of GPT-3. In addition, our analysis unveilsnew insights, with detailed rationales provided by laypeople, e.g., that thecommonsense capabilities have been improving with larger models while mathcapabilities have not, and that the choices of simple decoding hyperparameterscan make remarkable differences on the perceived quality of machine text. Werelease our training material, annotation toolkit and dataset athttps://yao-dou.github.io/scarecrow/.

Is GPT-3 Text Indistinguishable from Human Text? Scarecrow: A Framework for Scrutinizing Machine Text