[Adapt] [Seminar] Commonsense Knowledge Evaluation

[vanessea]

Hi Adapters,


Commonsense knowledge (CSK) is receiving significant attention as it plays an important role in building robust AI systems. Prior works on acquiring CSK have compiled statements about concepts and their properties, but most works treat each statement in isolation from each other and use a single “confidence” score as the only quantitative evaluation measure.




In this seminar, I’ll introduce the paper “Joint Reasoning for Multi-Faceted Commonsense Knowledge”. The authors propose a “multi-facet” notion of scoring CSK, arguing that it is not sufficient to have just a single “confidence” score, but instead propose four distinct values: plausibility, typicality, remarkability, and salience. These values are codified into a series of constraints that conform to their natural language definitions, leading to an approach for extending existing CSKBs based on this new multi-faceted perspective.


Time: Wed 4:00pm
Venue: SEIEE 3-414

See you tomorrow!
Best wishes,
Ilena
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://cs.sjtu.edu.cn/pipermail/adapt/attachments/20201020/db6a9737/attachment.html>