Zengzhi Wang

I am a second-year master student at Text Mining Group, Nanjing University of Science & Technology, fortunately supervised by Prof. Rui Xia and Assoc. Prof. Jianfei Yu. I earned my B.Eng. degree at the School of Computer Science & Engineering, Wuhan Institute of Technology.

My research interests lie at the intersection of natural language processing and machine learning. I did research in the Fine-Grained (i.e., Aspect-Based) Sentiment Analysis and Opinion Mining since 2021.

Currently, I am very interested in Large Language Models, Complex Reasoning, Instruction Learning, Embodied AI (in the long term), etc.

I’m actively looking for a Ph.D. position in the 2024 Fall. Free free to contact with me for any form of collaboration.

Email  /  Google Scholar  /  Twitter  /  Github

• 2023-04-11 – We release a technical report about the sentiment understanding abilities of ChatGPT. [Paper] [Code]

• 2022-04-07 – Our Instruction-Tuning Reading List Repo receives 200+ stars!

• 2022-04-05 – 1 paper (FS-ABSA) accepted by SGIR'23!

• 2022-11-19 – Updated this homepage. No good news. I got rejected twice recently.

We introduce a large-scale Multi-Element Multi-Domain dataset (MEMD) that covers the four elements across five domains for ABSA research.

Can ChatGPT really understand the opinions, sentiments, and emotions contained in the text? Our evaluation on 5 sentiment analysis tasks and 18 benchmark datasets shows that ChatGPT can function as a versatile and reliable sentiment analyzer, even without domain-specific training.

We argue that domain gap and objective gap hinder the transfer of knowledge from PLMs to ABSA tasks. To this end, we introduce a simple yet effective framework called FS-ABSA, which involves domain-adaptive pre-training and textinfilling fine-tuning.

A general-purpose ABSA framework based on multi-task instruction tuning that can uniformly model various ABSA tasks and capture the inter-task dependency with multi-task learning.

Last modified in Nov. 2022. Design and source code from Jon Barron.