Classification of Obsessive-Compulsive Disorder Symptoms in Arabic Tweets Using Machine Learning and Word Embedding Techniques

Abstract

Obsessive-Compulsive Disorder (OCD) is a mental health condition that is characterized by persistent and intrusive thoughts, images, or impulses (obsessions), as well as the presence of repetitive behaviors or mental acts (compulsions) that are aimed at reducing anxiety. These behaviors are typically rigid and performed according to specific rules; furthermore, they can be time-consuming and cause significant distress or impairment in daily functioning. Detecting the symptoms of OCD could help individuals become aware of them and seek a medical diagnosis. People increasingly rely on social media such as Twitter to express and disclose their feelings and thoughts. Although researchers have researched OCD in English, there is no substantial work in this domain concerning Arabic tweets. Therefore, this research proposes investigating and detecting OCD in Arabic tweets using Machine Learning (ML) and word embedding techniques. First, we obtained tweets via web scraping and manually annotated the data by involving medical professionals. Secondly, we conducted exploratory data analysis on the textual data and emojis used to find their correlation. Furthermore, we focused on the quality of word representation. Efficient word representation approaches (word embeddings) combined with recent ML models have shown reasonable progress on text classification tasks. We trained our classification model using the Arabic version of fastText. The proposed models were also tested on our dataset. The analysis indicated that utilizing fastText as a word embedding technique is a particularly promising approach.