China's four decades of rural-to-urban labor migration has created 36 million migrant youth, who encounter more adversity than their urban peers due to educational and socioeconomic exclusion. This study focuses on the psychological resilience of these migrant youth since it indicates the ability or success to cope with adversity. We examine psychological resilience indicators based on migrant youth language patterns extracted in a multidisciplinary way. Our research questions are: whether Chinese migrant youths' language patterns indicate their resilience-related characteristics, and how effectively can three language analysis approaches (user-defined dictionary, feature extraction, and word co-occurrence) identify these indicators, which may be further applied as machine learning features for psychological resilience level estimation. Students in a middle school for migrant youth in Shenzhen, China were recruited as initial participants, and their resilience levels were rated by the Chinese version of Connor-Davidson Resilience Scale. These youth then wrote one week of diary entries. The high- (n = 37) and low-resilience (n = 45) participants' writings were analyzed through the three language analysis approaches. The results suggest that Chinese migrant youth in this study present distinctive language patterns that closely relate to resilience indicators (positive emotion, self-efficacy, self-esteem, and positive affect), which can be further employed in differentiating resilience level (high vs. low) as machine learning features with satisfactory prediction accuracy. Our findings suggest a new interdisciplinary approach for effective psychological resilience identification among migrant youth in China and other high-risk youth populations. © 2019 Elsevier Ltd

Chinese Language analysis approaches Multidiscipline Psychological resilience Language patterns migrant youth

China writing human controlled study language self esteem migrant Adolescent male female prediction juvenile middle school clinical article Article psychological resilience human experiment Connor Davidson resilience scale feature extraction