From exact matching to semantic matching: Using neural models for ranking

Speaker: Zhenghao Liu
         Tsinghua University

Title:  "From exact matching to semantic matching:
         Using neural models for ranking"

Date:   Monday, 12 April 2021

Time:   4:00pm - 5:00pm

Zoom link:
https://hkust.zoom.us/j/95482806841?pwd=NXpLVDVVOWhiN09EcmM1SndmNVU2Zz09

Meeting ID:     954 8280 6841
Passcode:       210412


Abstract:

Early work in Information Retrieval (IR) usually employs traditional IR
models, such as BM25, and mainly focuses on the exacting match. With the
development of the deep neural network, IR tasks also enjoy the benefit of
neural models. Neural IR models help to deal with the vocabulary mismatch
problem from traditional IR models and focus more on semantic matching.
However, there are also some work also argues that if neural network
really works for IR tasks, especially for few-shot ranking tasks. This
talk will introduce widely used neural ranking models used in IR and how
to take advantage of neural IR models.


*****************
Biography:

Zhenghao Liu is a Ph.D. candidate at Tsinghua University. His research
interest is information retrieval and question answering. He had published
several papers on ACL, EMNLP and WWW.