CLPsych 2015 Shared Task: Depression and PTSD on Twitter

Abstract

This paper presents a summary of the Computational Linguistics and Clinical Psychology (CLPsych) 2015 shared and unshared tasks. These tasks aimed to provide apples-to-apples comparisons of various approaches to modeling language relevant to mental health from social media. The data used for these tasks is from Twitter users who state a diagnosis of depression or post traumatic stress disorder (PTSD) and demographically-matched community controls. The unshared task was a hackathon held at Johns Hopkins University in November 2014 to explore the data, and the shared task was conducted remotely, with each participating team submitted scores for a held-back test set of users. The shared task consisted of three binary classification experiments: (1) depression versus control, (2) PTSD versus control, and (3) depression versus PTSD. Classifiers were compared primarily via their average precision, though a number of other metrics are used along with this to allow a more nuanced interpretation of the performance measures.

Natural language processing in mental health applications using non-clinical texts

“…The various works of De Choudhury et al, and the participants of the shared task hosted by Coppersmith et al (2015) all attempt to make clinical diagnoses (for depression, posttraumatic stress and postpartum depression) from social media data. Homan et al (2014b) and Masuda et al (2013) aim to identify people who are in current distress or are contemplating suicide.…”

“…Another shared task was a 2015 Computational Linguistics and Clinical Psychology (CLPsych) shared task (Coppersmith et al 2015). The data used for the task, and a hackathon held at John Hopkins University, consisted of anonymized Tweets written by 1,746 users who stated a diagnosis of depression or posttraumatic stress disorder (PTSD), with demographically-matched community controls.…”

“…A classifier was built that could from the prepartum data alone predict the mothers who would later exhibit these dramatic changes with an accuracy of 71%. Coppersmith et al (2015) host a shared task that also attempts to make clinical diagnoses from Twitter data; in this case, for depression and posttraumatic stress disorder. Their work, and that of the task participants, is described in Section 4.4.…”

Multitask Learning for Mental Health Conditions with Limited Social Media Data

“…Written text carries implicit information about the author, a relationship that has been exploited in natural language processing (NLP) to predict author characteristics, such as age (Goswami et al, 2009;Rosenthal and McKeown, 2011;Nguyen et al, 2011;Nguyen et al, 2014), gender (Sarawgi et al, 2011;Ciot et al, 2013;Liu and Ruths, 2013;Alowibdi et al, 2013;Volkova et al, 2015;Hovy, 2015), personality and stance (Schwartz et al, 2013b;Schwartz et al, 2013a;Volkova et al, 2014;Plank and Hovy, 2015;Preoţiuc-Pietro et al, 2015), or occupation (Preotiuc-Pietro et al, 2015a;Preoţiuc-Pietro et al, 2015b). The same signal has also been effectively used to predict mental health conditions, such as depression (Coppersmith et al, 2015b;Schwartz et al, 2014), suicidal ideation (Coppersmith et al, 2016;Huang et al, 2015), schizophrenia (Mitchell et al, 2015) or post-traumatic stress disorder (PTSD) (Pedersen, 2015), often more accurately than by traditional diagnoses.…”

Using contextual information for automatic triage of posts in a peer-support forum

“…In this context, text mining tools are emerging as a powerful channel to study and detect the mental state of the writers (Calvo and Mac Kim, 2013;Bedi et al, 2015Bedi et al, , 2014De Choudhury et al, 2013a,b;Coppersmith et al, 2015). In particular, there is a greater interest in the study and detection of suicidal ideation in texts coming from social networks.…”