@Inbook{Kulkarni2023,
author="Kulkarni, Hrishikesh
and MacAvaney, Sean
and Goharian, Nazli
and Frieder, Ophir",
editor="Shaban-Nejad, Arash
and Michalowski, Martin
and Bianco, Simone",
title="Knowledge Augmentation for Early Depression Detection",
bookTitle="Artificial Intelligence for Personalized Medicine: Promoting Healthy Living and Longevity",
year="2023",
publisher="Springer Nature Switzerland",
address="Cham",
pages="175--191",
abstract="Individuals continue to share their mental health concerns on social media, providing an avenue to rapidly detect those potentially in need of assistance. While users of immediate need can be recognized with relative ease, early-stage disorder users in the boundary region pose a greater challenge to detect. The minimal posting histories of such users further complicate proceedings. However, these same boundary region users would benefit greatly from timely treatment; hence, detecting their mental health status is of utmost need. Additionally, pointers to identify the type of depression could be of great help. Augmenting knowledge for low posting users can help to solve this problem. We propose an NLP based method `STBound' that intelligently determines the optimal region for knowledge augmentation. It answers three crucial questions: when?, for whom? and how much? to augment---to resolve this imbroglio. Our proposed selective knowledge augmentation method contributes to early depression detection performance improvement by an average of 11.9{\%} in F1 score. Further, this approach shows promising performance enhancement of 12.1{\%} in F1 score for the critical task of separating these boundary region users with bipolar depression. STBound identifies those depressed users in the boundary region who would otherwise go unidentified.",
isbn="978-3-031-36938-4",
doi="10.1007/978-3-031-36938-4_14",
url="https://doi.org/10.1007/978-3-031-36938-4_14"
}