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

The field of natural language understanding has experienced exponentialprogress in the last few years, with impressive results in several tasks. Thissuccess has motivated researchers to study the underlying knowledge encoded bythese models. Despite this, attempts to understand their semantic capabilitieshave not been successful, often leading to non-conclusive, or contradictoryconclusions among different works. Via a probing classifier, we extract theunderlying knowledge graph of nine of the most influential language models ofthe last years, including word embeddings, text generators, and contextencoders. This probe is based on concept relatedness, grounded on WordNet. Ourresults reveal that all the models encode this knowledge, but suffer fromseveral inaccuracies. Furthermore, we show that the different architectures andtraining strategies lead to different model biases. We conduct a systematicevaluation to discover specific factors that explain why some concepts arechallenging. We hope our insights will motivate the development of models thatcapture concepts more precisely.

Inspecting the concept knowledge graph encoded by modern language models