Some Reflections on Drawing Causal Inference with Texts

Speaker: Jiayao Zhang
         University of Pennsylvania

Title:  "Some Reflections on Drawing Causal Inference with Texts"

Date:   Thursday, 24 November 2022

Time:   10:00am - 11:00am

Venue:  Room 4502 (via lift 25/26), HKUST


Abstract:

Textual data is ubiquitous in various domains of modern machine learning
and data science. In many practical scenarios, one wishes to establish
either causal links among semantic meanings of texts or drawing
conclusions on causal queries that involve textual data. In this talk, we
discuss several research problems in this regard, including commonsense
causality reasoning and drawing causal queries involving textual data. We
will give an overview of different approaches and frameworks for these
problems, including our ROCK framework for the CCR task that is based on
the potential-outcomes framework. We finish the talk by discussing
outstanding problems and avenues for future work. This talk is based on
https://arxiv.org/abs/2202.00436, https://arxiv.org/abs/2202.00848, and
some ongoing work.


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

Jiayao Zhang graduated from HKU with a Bachelors in Engineering (Computer
Science) in 2019 and is currently a Ph.D. candidate at the University of
Pennsylvania advised by Dan Roth and Weijie Su. Jiayao's research interest
spans machine learning theories, commonsense reasoning in natural language
processing, causal inference, and algorithmic fairness. Jiayao has been an
intern in the ML&Forecasting Team under the AWS AI Labs since May 2022.
More information can be found at https://www.jiayao-zhang.com.