Watch the video here of my talk at Google's NLP Reading Group, where I discuss my paper on subjectivity and disagreement. |
Here I examine conversation acts and intents of witnesses in U.S. Congressional hearings, but with the twist that these labels are subjective. We create a model to predict these labels, and find that taking into account the annotator's sentiment towards the witness does help, consistent with quantitative and qualitative analyses of our dataset. Our paper is here. Download the data and reproduce our experiments on our repo. |
I'm presenting my first Abridge research collaboration at the AAAI 2021 Workshop: Trustworthy AI for Healthcare. We delve into the performance of the classifier that picks out key pieces of doctor-patient conversations for helping users stay on top of their health. We investigate how well our classifier performs across different socio-demographic groups to understand any disparities in how this benefit is being allocated. Our paper is on arvix here. |
I'm now a Research Scientist at Abridge. Excited to join a team committed to enabling patients to take control over their health care. |
If you're headed to Florence, come by my poster @Arsenale on Monday (10:30am-12:10pm) to chat about how you can (not) use attention to induce discourse structure. paper |
If you're at NAACL, come see me on June 6th 11:30AM @ Nicollet D3. I'll be talking about the successes and failures of using News RST segmenters on Medical data, and presenting a small corpus. paper |
I'll be interning with Google San Francisco, working with Praveen Paritosh. |
I'll be spending the summer in Palo Alto, CA, as an intern at Salesforce Research, working with James Bradbury. |
I completed my Master's in Linguistics and am now one step closer to a PhD! |
I am a Research Scientist in Natural Language Processing at Abridge.
Prior to joining Abridge, I received my Ph.D. in Linguistics from UT Austin in 2020, where I was advised by Dr. Katrin Erk and Dr. Jessy Li, and was also a member of the TAUR Lab headed by Dr. Greg Durrett.
My current research interests include conversation analysis and information extraction in the medical domain.
During my graudate studies, I was interested in computational discourse analysis and how to capture the ambiguity and subjectivity of discourse in dialogue.
E. Ferracane, G. Durrett, J.J. Li, K. Erk. Did they answer? Subjective acts and intents in conversational discourse. NAACL, 2021. [bib] [code] [poster] [slides] [video]
E. Ferracane, S. Konam. Towards Fairness in Classifying Medical Conversations into SOAP Sections. AAAI 2021 Workshop: Trustworthy AI for Healthcare, 2021. [bib] [poster]
E. Ferracane, G. Durrett, J.J. Li, K. Erk. Evaluating Discourse in Structured Text Representations. ACL, 2019. [bib] [code] [poster]
E. Ferracane, T. Page, J.J. Li, K. Erk. From News to Medical: Cross-domain Discourse Segmentation. NAACL DISRPT Workshop, 2019. [bib] [code] [slides]
E. Ferracane, S. Wang, R. Mooney. Leveraging discourse information effectively for authorship attribution. IJCNLP, 2017. [bib] [code] [slides] [video]
E. Ferracane, I. Marshall, B. Wallace, K. Erk. Leveraging coreference to identify arms in medical abstracts: An experimental study. EMNLP LOUHI Workshop, 2016. [bib] [code] [slides]
[pdf]
email: elisa AT ferracane DOT com