Quality Alchemist (品質煉金術師): PolyU Computational Linguistics Summit in the Era of LLM

The summit named “Computational Linguistics Summit in the Era of Large Language Models cum International symposium on Collaborative Innovations between The Hong Kong Polytechnic University and The China Computer Federation” organized by COMP Dept., PolyU on 22^nd Aug 2024. Day 1 had two guest speakers and they are Prof. Kam Fai WONG and Prof. Guodon ZHOU.

The first guest speaker was Prof. Kam Fai WONG (CUHK) and his topic named “What is Digital Humanities?” In the beginning, Prof. Wong briefed the different between Computational Linguistics (CL) and Natural language processing (NLP). CL is deeply rooted in both linguistics and computer science, whereas NLP is more closely aligned with AI and machine learning; indicating that CL is more concerned with the theoretical understanding of language, and NLP is focused on practical applications.

Then Prof. KF Wong shared the conference of ACL 2024 by Association for Computational Linguistics (ACL) that he just attended. He firstly mentioned three keynotes including Barbara Plank, Subbarao Kambhampati and Sunita Sarawagi. And then he shared the theme paper award entitled “OLMo Accelerating the Science of Language Models”.

After that Prof. Wong explained what is digital humanities? Digital humanities is an interdisciplinary field that applies digital tools and methods to enhance the research, teaching, and dissemination of knowledge in the humanities.

There are two paths for research to study LLMs that AI act as human or AI as itself. He also stated impacts to Humanities including language, history, philosophy (ethics), art and education, etc.

Finally, Prof. KF Wong introduced methods which developed by his research team, focusing on enhancing the forgettability, reliability, adaptability, multiplicity, and explainability (FRAME) of large models. Where:

Forgettability – some of data would be obsoleted (e.g. person die) that how to take it out.

Reliability – evaluation to make sure reliable

Adaptability – realize his/her behavior, emotion

Multiplicity – multi-model such as prompt

Explainability – popular problem in deep learning

After finished Q&A session, organizer presented souvenir to Prof. Wong.

Prof. Guodon ZHOU (周國棟) (Soochow University) and his presentation entitled “Language Mechanism” (語言機制). Firstly, he explained that language mechanisms refer to the natural processes involved in human language use, while language models are computational structures designed to imitate or facilitate language-related tasks.

Then he further explained the language mechanisms through NLP that separated into three formalism (形式主義), functionalism (功能主義) and cognitivism (認知主義).

The difference between linguistic learning (LL) and natural language　understanding (NL) that LL pertains to the acquisition of language knowledge, whether by humans or machines, while NL focuses on enabling machines to comprehend and interpret human language meaningfully.