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

Deep-learning based Automatic Essay Scoring (AES) systems are being activelyused by states and language testing agencies alike to evaluate millions ofcandidates for life-changing decisions ranging from college applications tovisa approvals. However, little research has been put to understand andinterpret the black-box nature of deep-learning based scoring algorithms.Previous studies indicate that scoring models can be easily fooled. In thispaper, we explore the reason behind their surprising adversarial brittleness.We utilize recent advances in interpretability to find the extent to whichfeatures such as coherence, content, vocabulary, and relevance are importantfor automated scoring mechanisms. We use this to investigate theoversensitivity i.e., large change in output score with a little change ininput essay content) and overstability i.e., little change in output scoreswith large changes in input essay content) of AES. Our results indicate thatautoscoring models, despite getting trained as "end-to-end" models with richcontextual embeddings such as BERT, behave like bag-of-words models. A fewwords determine the essay score without the requirement of any context makingthe model largely overstable. This is in stark contrast to recent probingstudies on pre-trained representation learning models, which show that richlinguistic features such as parts-of-speech and morphology are encoded by them.Further, we also find that the models have learnt dataset biases, making themoversensitive. To deal with these issues, we propose detection-based protectionmodels that can detect oversensitivity and overstability causing samples withhigh accuracies. We find that our proposed models are able to detect unusualattribution patterns and flag adversarial samples successfully.

AES Systems Are Both Overstable And Oversensitive: Explaining Why And Proposing Defenses