Publications

A Prioritization Model for Suicidality Risk Assessment

Published in Association for Computational Linguistics (ACL), 2020

We reframe suicide risk assessment from social media as a ranking problem whose goal is maximizing detection of severely at-risk individuals given the time available. We introduce a well founded evaluation paradigm, and demonstrate using an expert-annotated test collection that meaningful improvements over plausible cascade model baselines can be achieved using an approach that jointly ranks individuals and their social media posts. Read more

Download here

Assigning Medical Codes at the Encounter Level by Paying Attention to Documents

Published in Machine Learning for Health Workshop (ML4H) at NeurIPS, 2019

We introduce encounter-level document attention networks, which use hierarchical attention to explicitly take the hierarchical structure of encounter documentation into account. Experimental evaluation demonstrates improvements in coding accuracy as well as facilitation of human reviewers in their ability to identify which documents within an encounter play a role in determining the encounter level codes. Read more

Download here

Unsupervised System Combination for Set-based Retrieval with Expectation Maximization

Published in International Conference of the Cross-Language Evaluation Forum for European Languages (CLEF), 2019

This paper presents a set-generating unsupervised system combination framework that draws inspiration from evaluation techniques in sparse data settings. It argues for the existence of a duality between evaluation and system combination, and then capitalizes on this duality to perform unsupervised system combination. Read more

Download here

Expert, Crowdsourced, and Machine Assessment of Suicide Risk via Online Postings

Published in Proceedings of the Fifth Workshop on Computational Linguistics and Clinical Psychology: From Keyboard to Clinic (CLPsych), 2018

We report on the creation of a dataset for studying assessment of suicide risk via online postings in Reddit. Evaluation of risk-level annotations by experts yields what is, to our knowledge, the first demonstration of reliability in risk assessment by clinicians based on social media postings. Read more

Download here