Spring 2023 Schedule

Location: Hybrid - Buchanan 126, and the zoom link below

Time: Wednesdays at 10:30am, Mountain Time

Zoom link: https://cuboulder.zoom.us/j/97014876908

Date	Title
01/18/23	Cancelled, due to weather
01/25/2023	Planning, introductions, welcome!
02/01/2023	TBD
02/08/2023	TBD
02/15/2023	Diego Garcia Moved to 03/01, due to weather
02/22/2023	Ashis Kumer Biswas (invited speaker; University of Colorado Denver) Abhidip Bhattacharyya: Presentation
03/01/2023	Diego Garcia
03/07/2023	Marjorie Freedman, ISI, Wikidata as an IE Ontology
03/08/2023	CLASIC Open House
03/15/2023	Role-Playing Paper-Reading: Decomposing and Recomposing Event Structure (https://tinyurl.com/p6mb7b7t)
03/22/2023	Ananya Ganesh: prelim
03/29/2023	Spring break -- no meeting!
04/05/2023	Kyle Gorman (invited speaker; City University of New York)
04/12/2023	Abteen Ebrahimi: prelim
04/19/2023	Jon Cai: proposal defense CANCELLED
04/26/2023	Elizabeth Spaulding: proposal defense Strand 1 iSAT Research - Understanding and Facilitating Collaborations - Jie Cao, Jon Cai, Ananya Ganesh, Martha Palmer
05/03/2023	Sameer Pradhan, GRAIL—Generalized Representation and Aggregation of Information Layers MOVED TO FALL 2023
05/17/2023	Skatje Myers, Practice Talk for Thesis Defense - Adapting Semantic Role Labeling to New Genres and Languages Abstract: Semantic role labeling (SRL) is the identification of semantic predicates and their participants within a sentence, which is vital for deeper natural language understanding. Current SRL models require annotated text for training, but this is unavailable in many domains and languages. We explore two different ways of reducing the annotation required to produce effective SRL models: 1) using active learning to target only the most informative training instances and 2) leveraging parallel sentences to project SRL annotations from one language into the target language.
05/18/2023	11am-1pm MT: Skatje Myers, REAL Thesis Defense - Adapting Semantic Role Labeling to New Genres and Languages. Abstract: Semantic role labeling (SRL) is the identification of semantic predicates and their participants within a sentence, which is vital for deeper natural language understanding. Current SRL models require annotated text for training, but this is unavailable in many domains and languages. We explore two different ways of reducing the annotation required to produce effective SRL models: 1) using active learning to target only the most informative training instances and 2) leveraging parallel sentences to project SRL annotations from one language into the target language.