ACQUIRING AND MODELLING ABSTRACT COMMONSENSE KNOWLEDGE VIA CONCEPTUALIZATION

MPhil Thesis Defence


Title: "ACQUIRING AND MODELLING ABSTRACT COMMONSENSE KNOWLEDGE VIA 
CONCEPTUALIZATION"

By

Miss Mutian HE


Abstract

Conceptualization, or viewing things and events as instances of abstract 
concepts in mind, and making inferences based on that, is a vital 
component in human intelligence for commonsense reasoning. Although recent 
artificial intelligence has made progress in acquiring and modelling 
commonsense thanks to the large neural language models and commonsense 
knowledge graphs (CKGs), conceptualization is yet to be introduced, making 
them ineffective to cover knowledge about countless diverse entities in 
the real world.

To address the problem, we thoroughly study the possible role of 
conceptualization in commonsense reasoning, and formulate a framework to 
replicate human conceptual induction from acquiring abstract knowledge 
about abstract concepts. Aided by Probase, We develop tools for 
contextualized concept identification, linking, and abstraction on ATOMIC, 
a large-scale human annotated CKG. We annotate a dataset for the validity 
of abstractions for ATOMIC on both event and triple level, and train a set 
of neural models to generate and discriminate abstract knowledge. xi Based 
on these components, a pipeline to acquire abstract knowledge is built. A 
large abstract CKG upon ATOMIC is then induced, ready to be instantiated 
for inferences over unseen entities. Furthermore, experiments show that 
injecting abstract triples is helpful in commonsense modelling.


Date:  			Thursday, 2 June 2022

Time:			10:00am - 12:00noon

Zoom Meeting:
https://hkust.zoom.us/j/98952940855?pwd=bHJtc1JPeWdUTTNFUjVvYnRiL3U0UT09

Committee Members:	Dr. Yangqiu Song (Supervisor)
 			Prof. Xiaofang Zhou (Chairperson)
 			Dr. Xiaojuan Ma


**** ALL are Welcome ****