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

Despite achieving state-of-the-art results in nearly all Natural LanguageProcessing applications, fine-tuning Transformer-based language models stillrequires a significant amount of labeled data to work. A well known techniqueto reduce the amount of human effort in acquiring a labeled dataset is\textit{Active Learning} (AL): an iterative process in which only the minimalamount of samples is labeled. AL strategies require access to a quantifiedconfidence measure of the model predictions. A common choice is the softmaxactivation function for the final layer. As the softmax function providesmisleading probabilities, this paper compares eight alternatives on sevendatasets. Our almost paradoxical finding is that most of the methods are toogood at identifying the true most uncertain samples (outliers), and thatlabeling therefore exclusively outliers results in worse performance. As aheuristic we propose to systematically ignore samples, which results inimprovements of various methods compared to the softmax function.

To Softmax, or not to Softmax: that is the question when applying Active Learning for Transformer Models