Talk by Kenneth Church at NUS on 30 May 2024

[Hwee Tou Ng]

You are cordially invited to a seminar at the NUS Advanced Research and Technology Innovation Centre (ARTIC).

 

Speaker: Professor Kenneth Church, Northeastern University in California

 

Date: 30 May 2024

Time: 2pm – 3pm

Venue: COM3 02-59 Meeting Room 20, NUS School of Computing

 

Chairperson: Professor Hwee Tou Ng, National University of Singapore

 

Title: Better Together: Text + Context

 

Abstract:

 

Graph learning has applications in web search (Page Rank), Product Search (Amazon), Biology, Finance and Traffic Analysis for defense.  We will focus on applications in Academic Search because the data is less sensitive and more available.  We argue that Content-based filtering (CBF) and Graph-based methods (GB) complement one another. The scientific literature can be viewed as a conversation between authors and the audience. CBF uses abstracts to infer authors' positions, and GB uses citations to infer responses from the audience. In this talk, we describe nine differences between CBF and GB, as well as synergistic opportunities for hybrid combinations. Two embeddings will be used to illustrate these opportunities: (1) Specter, a CBF method based on BERT-like deepnet encodings of abstracts, and (2) ProNE, a GB method based on spectral clustering of more than 200M papers and 2B citations from Semantic Scholar (S2).   Resources are available: APIs, a website, and several embeddings of S2 papers.

 

Biography:

 

Ken Church is a Professor of the Practice at Northeastern University in California. He works in computational linguistics and many related areas including web search, large language models, lexicography and compression.  He was an early advocate of empirical methods and a founder of the Conference on Empirical Methods in Natural Language Processing (EMNLP).  Honors: Fellow of ACM, ACL, AT&T, Baidu. Web Page: https://kwchurch.github.io/

 

To register for this seminar, please email Ms Alexandra Chua: cfka...@nus.edu.sg