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

Progress on commonsense reasoning is usually measured from performanceimprovements on Question Answering tasks designed to require commonsenseknowledge. However, fine-tuning large Language Models (LMs) on these specifictasks does not directly evaluate commonsense learned during pre-training. Themost direct assessments of commonsense knowledge in pre-trained LMs arearguably cloze-style tasks targeting commonsense assertions (e.g., A pen isused for [MASK].). However, this approach is restricted by the LM's vocabularyavailable for masked predictions, and its precision is subject to the contextprovided by the assertion. In this work, we present a method for enriching LMswith a grounded sense inventory (i.e., WordNet) available at the vocabularylevel, without further training. This modification augments the predictionspace of cloze-style prompts to the size of a large ontology while enablingfiner-grained (sense-level) queries and predictions. In order to evaluate LMswith higher precision, we propose SenseLAMA, a cloze-style task featuringverbalized relations from disambiguated triples sourced from WordNet, WikiData,and ConceptNet. Applying our method to BERT, producing a WordNet-enrichedversion named SynBERT, we find that LMs can learn non-trivial commonsenseknowledge from self-supervision, covering numerous relations, and moreeffectively than comparable similarity-based approaches.

Probing Commonsense Knowledge in Pre-trained Language Models with Sense-level Precision and Expanded Vocabulary