Machine Learning and Natural Language Processing in Mental Health: Systematic Review

Background Machine learning systems are part of the field artificial intelligence that automatically learn models from data to make better decisions. Natural language processing (NLP), by using corpora and approaches, provides good performance in statistical tasks, such as text classification or sentiment mining. Objective The primary aim this systematic review was summarize characterize, methodological technical terms, studies used machine NLP techniques for mental health. secondary consider potential use these methods health clinical practice Methods This follows PRISMA (Preferred Reporting Items Systematic Review Meta-analysis) guidelines is registered with PROSPERO (Prospective Register Reviews; number CRD42019107376). search conducted 4 medical databases (PubMed, Scopus, ScienceDirect, PsycINFO) following keywords: learning, mining, psychiatry, health, disorder. exclusion criteria were follows: languages other than English, anonymization process, case studies, conference papers, reviews. No limitations on publication dates imposed. Results A total 327 articles identified, which 269 (82.3%) excluded 58 (17.7%) included review. results organized through a qualitative perspective. Although had heterogeneous topics methods, some themes emerged. Population could be grouped into 3 categories: patients databases, who came emergency room, social media users. main objectives extract symptoms, classify severity illness, compare therapy effectiveness, provide psychopathological clues, challenge current nosography. Medical records 2 major sources. With regard used, preprocessing standard unique identifier extraction dedicated texts. Efficient classifiers preferred rather transparent functioning classifiers. Python most frequently platform. Conclusions have been highly topical issues medicine recent years may considered new paradigm research. However, processes tend confirm hypotheses developing entirely information, only one category population (ie, users) an imprecise cohort. Moreover, language-specific features can improve their extension should more closely investigated. useful information unexplored patients’ daily habits usually inaccessible care providers). Before considering It additional tool care, ethical remain discussed timely manner. offer multiple perspectives research but also tools support practice.