Contextual Word Embeddings and Polysemy in Hierarchical Topic Modeling

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering

Final Year Thesis Oral Defense

Title: "Contextual Word Embeddings and Polysemy in Hierarchical Topic 
        Modeling"

by

WANG Zhiwei

Abstract:

Hierarchical topic modeling (HTD) is a variant of topic modeling task 
which has attracted researchers' attention in recent years. Hierarchical 
latent tree analysis (HLTA) is a recently proposed method for HTD that has 
shown superior performance than state-of-the- art methods. However, the 
model of HLTA has a tree structure and cannot represent different meanings 
of a word appropriately. We proposed a method using contextual word 
embeddings to add context information to the text corpus before running 
HLTA. Some selected words are replaced by their specific word senses in 
the corpus. The empirical evaluation shows that the proposed method 
outperforms the original HLTA on the test dataset.


Date            : 8 May 2021 (Saturday)

Time            : 16:00-16:40

Zoom Link:
https://hkust.zoom.us/j/92133587886?pwd=N3orUFNBNFpSVjZiV1JTM0Y5TVNZZz09

Meeting ID      : 921 3358 7886
Passcode        : 663031

Advisor         : Prof. ZHANG Nevin Lianwen

2nd Reader      : Dr. SONG Yangqiu