03861nas a2200469 a 4500 20260817173419.0 240229s2022 th u m tt 000 a eng d .b12421662 AIT Thesis no.DSAI-22-04 Chanapa Pananookooln Comparing selective masking methods for depression detection in social media Pathum Thani, Thailand : Asian Institute of Technology, 2022 47 leaves : ill. Thesis ; no. DSAI-22-04 A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Data Science and Artificial Intelligence, School of Engineering and Technology Thesis (M. Sc.) - Asian Institute of Technology, 2022 Identifying those at risk for depression is a crucial issue in which social media provides an excellent platform for examining the linguistic patterns of depressed individuals. A significant challenge in a depression classification problem is ensuring that the predic tion model is not overly dependent on keywords, such that it fails to predict when key words are unavailable. One promising approach is masking, i.e., by masking important words selectively and asking the model to predict the masked words, the model is forced to learn the context rather than the keywords. This study evaluates seven masking tech niques, such as random masking, log-odds ratio, and the use of attention scores. In ad dition, whether to predict the masked words during pretraining or fine-tuning phase was also examined. Last, six class imbalance ratios were compared to determine the robust ness of the masked selection methods. Key findings demonstrated that selective masking generally outperforms random masking in terms of classification accuracy. In addition, the most accurate and robust models were identified. Our research also indicated that re constructing the masked words during the pre-training phase is more advantageous than during the fine-tuning phase. Further discussion and implications were made. This is the first study to comprehensively compare masking selection methods, which has broad implications for the field of depression classification and the general NLP. Our code can be found in: https://github.com/chanapapan/Depression-Detection Social media Data processing Machine learning Neural networks (Computer science) Chaklam Silpasuwanchai, Chairperson Dailey, Matthew N., Examination Committee Mongkol Ekpanyapong, Examination Committee His Majesty the King{u2019}s Scholarships (Thailand), Scholarship Donor Asian Institute of Technology. Thesis ; no. DSAI-22-04 Full-Text http://203.159.5.9/ait-thesis/Viewer/viewer.php?id=B20417 .b12421662 mnait a 250428 0 240313 m a a 0 mnarc mnarc mnait mnait 67 40 20 Barcode : - CREATED : 2024-02-29 RECORD # : i13487875 LPATRON : 0 LCHKIN : - # RENEWALS : 0 # OVERDUE : 0 IUSE3 : 0 TOT CHKOUT : 0 TOT RENEW : 0 Barcode : 30050120899066 CREATED : 2025-07-03 RECORD # : i13537118 LPATRON : 0 LCHKIN : - # RENEWALS : 0 # OVERDUE : 0 IUSE3 : 0 TOT CHKOUT : 0 TOT RENEW : 0 Barcode : 30050120425151 CREATED : 2025-08-04 RECORD # : i1354293x LPATRON : 0 LCHKIN : - # RENEWALS : 0 # OVERDUE : 0 IUSE3 : 0 TOT CHKOUT : 0 TOT RENEW : 0 Barcode : 30050120425144 CREATED : 2025-08-04 RECORD # : i13542941 LPATRON : 0 LCHKIN : - # RENEWALS : 0 # OVERDUE : 0 IUSE3 : 0 TOT CHKOUT : 0 TOT RENEW : 0 19836 19836 0 0 0 0 MAIN MAIN mnarc 2026-08-17 0 AIT Thesis no.DSAI-22-04 2026-08-17 17:34:19 2026-08-17 67 0 0 0 0 MAIN MAIN mnarc 2026-08-17 0 AIT Thesis no.DSAI-22-04 30050120899066 2026-08-17 17:34:19 1 2026-08-17 40 0 0 0 0 MAIN MAIN mnait 2026-08-17 50.00 0 AIT Thesis no.DSAI-22-04 30050120425151 2026-08-17 17:34:19 1 2026-08-17 20 0 0 0 0 MAIN MAIN mnait 2026-08-17 50.00 0 AIT Thesis no.DSAI-22-04 30050120425144 2026-08-17 17:34:19 2 2026-08-17 20