Social Media NLP Analysis for Mental Health Predictive Intervention

!

We look into social media to predict mental health ailments and signal to practitioners to supply preventive care or initiate care giving.

Installation

Follow requirements.txt (spaCy has an extra step)

Language features

For the syntactic features, run:

import pandas as pd
import content
df = pd.read_pickle("suicidewatch-sample.pkl")
df = content.addSyntacticFeatures(df)

For the affection features, run:

import afinnsenti
import labmt
df['text'] = df.apply(content.getTextFromRecord, axis=1)
df = afinnsenti.addEmotionalFeature(df)
df = labmt.addEmotionalFeature(df)

Binary classification

import binaryClassification
binaryClassification.main()
rs = binaryClassification.readResults()

The complete output of the classification results is also stored as a dictionary in pickle format (file: ./data/combinations-10fold.pkl)

Wordclouds

Follow the link

Reference Papers

The Language of Mental Health Problems in Social Media, Computational Linguistics and Clinical Psychology

• George Gkotsis, Anika Oellrich, Tim Hubbard, Richard Dobson, Maria Liakata, Sumithra Velupillai and Rina Dutta. The Language of Mental Health Problems in Social Media, Computational Linguistics and Clinical Psychology (clpsych, NAACL 2016).
• paper
• supplement
• Reddit Datasets (comments, posts).