Evaluation and application of abstract meaning representation

PhD Thesis Proposal Defence


Title: "Evaluation and application of abstract meaning representation"

by

Miss Ziyi SHOU


Abstract:

Understanding the meaning of natural language has been a long-time goal in the
field of natural language processing. Meaning representation acts as a
connection between linguistic expressions and the underlying meaning of the
words used. Despite the advancements in vector-based encoding of meaning,
discrete and hierarchically structured semantic representations remain crucial
in natural language understanding because of their interpretability and
reliability. Abstract Meaning Representation (AMR) is a typical example of
meaning representation, which represents the meaning of a sentence as a
directed graph with concepts as labeled nodes and relations as directed edges.
The primary objective of our research is to investigate the potential
applications ofAMRin downstream natural language tasks.

This proposal first introduce an AMR metric for efficient similarity evaluation
and selection of high-performing AMR parsers. Once a parser is selected, it can
be used to parse sentences into AMR graphs. These graphs can then undergo a
series of modifications, resulting in a large dataset of paraphrased sentences.
The experiments show that the data augmented by our AMR-DA method is beneficial
for downstream tasks. Furthermore, we propose exploring other potential
research topics for AMR within the context of large language models.


Date:                   Friday, 24 November 2023

Time:                   4:00pm - 6:00pm

Venue:                  Room 4475
                        lifts 25/26

Committee Members:      Prof. Fangzhen Lin (Supervisor)
                        Prof. Ke Yi (Chairperson)
                        Dr. Junxian He
                        Dr. Dan Xu


**** ALL are Welcome ****