Suicide is a subjective and complex phenomenon with a low base rate. Data on suicide is usually collected from healthcare organizations, large-scale studies, or self reporting Crosby et al. (2011); Horowitz and Ballard (2009). These sources are often limited by sociocultural barriers, such as stigma and shame Crosby et al. (2011). Many researchers focus on research related to associating risk factors and suicidaility without considering theoretical models Nock et al. (2008). The know risk factors are demographics, previous suicide attempts, mental health concerns (i.e., depression, substance abuse, suicidal ideation, self-harm, or impulsivity), family history of suicide, interper- sonal conflicts (i.e., family violence or bullying), and mechanism,
…show more content…
Early work consisted of classifying movie reviews (Pang et al., 2002) and product reviews (Turney, 2002) as positive or negative. Pang et al. (2002) and Snyder and Barzilay (2007) were then able to predict the ratings (based on 4, 5 or 10 point scale) of movies and restaurant based on their reviews. To classify these sentiments: Na ̈ıve Bayes, Maximum Entropy and SVM are the popular machine learning algorithms. Unlike traditional classifiers, where the neutral class is ignored, classifiers predicting sentiment are known to benefit with the addition of the neutral class. This task has of sentiment analysis can also classify human mood such as angry, happy or sad. Read (2005) and Go et al. (2009) consider the usage of emoticons (such as :) and :-/) to predict sentiment of the text from Usenet newsgroups and Twitter using a dataset of positive and negative …show more content…
Choudhury et al. (2013) and Prieto et al. (2014) successfully use Twitter data to detect depression and other mental health conditions, and argue that Twitter text data is viable to capture individuals psychological state. Choudhury et al. (2013) employed crowdsourcing to obtain a set of Twitter user who are clinically depressed based on standard psychometric tests. The author then retrieved their social information for last year and extracted behavioral text features and network features to predict the on-set of depression. Prieto et al. (2014) attempts to predict health conditions such as flu, depression, pregnancy and eating disorders by extracting relevant tweets to each category using a set of regular expressions and then classifying these conditions using an SVM with mean f-measure of
Early work consisted of classifying movie reviews (Pang et al., 2002) and product reviews (Turney, 2002) as positive or negative. Pang et al. (2002) and Snyder and Barzilay (2007) were then able to predict the ratings (based on 4, 5 or 10 point scale) of movies and restaurant based on their reviews. To classify these sentiments: Na ̈ıve Bayes, Maximum Entropy and SVM are the popular machine learning algorithms. Unlike traditional classifiers, where the neutral class is ignored, classifiers predicting sentiment are known to benefit with the addition of the neutral class. This task has of sentiment analysis can also classify human mood such as angry, happy or sad. Read (2005) and Go et al. (2009) consider the usage of emoticons (such as :) and :-/) to predict sentiment of the text from Usenet newsgroups and Twitter using a dataset of positive and negative …show more content…
Choudhury et al. (2013) and Prieto et al. (2014) successfully use Twitter data to detect depression and other mental health conditions, and argue that Twitter text data is viable to capture individuals psychological state. Choudhury et al. (2013) employed crowdsourcing to obtain a set of Twitter user who are clinically depressed based on standard psychometric tests. The author then retrieved their social information for last year and extracted behavioral text features and network features to predict the on-set of depression. Prieto et al. (2014) attempts to predict health conditions such as flu, depression, pregnancy and eating disorders by extracting relevant tweets to each category using a set of regular expressions and then classifying these conditions using an SVM with mean f-measure of