在社交媒體中理解和應對低自尊表達是非常重要的，因為它直接影響個人的心理健康和整體生活質量。本研究提出了一種新的方法來確定推文中低自尊的例子，利用了零標記分類和來自ConceptNet的上下文信息的獨特組合。所提出的方法應用了多模態聚類和具有孿生網路的對比學習來指派概率分數，從而提高了對心理健康相關表達的分析準確性。利用ConceptNet對普遍知識的廣泛涵蓋增強了模型的上下文理解能力，提高了低自尊檢測的準確性。人工評估驗證了精確概率分數的有效性，展示了模型捕捉了先前被忽視的與低自尊相關的潛在語境的能力。實驗結果還展示了所提出方法相對於傳統分類模型的卓越性能，凸顯了對於更好理解和支持正在經歷低自尊的人之潛力。該研究為分析社交媒體中的心理健康表達打開了新的可能性，有助於改善健康和心理健康干預措施。

Understanding and addressing low self-esteem expressions in social media is of significant importance, as it directly impacts an individual's mental well-being and overall quality of life. This research presents a novel methodology for determining low self-esteem instances in tweets, utilizing a unique combination of zero-shot classification and contextual information from ConceptNet. The proposed approach applies multimodal clustering and contrastive learning with a Siamese network to assign probability scores, enabling the analysis of mental health-related expressions with enhanced accuracy. Leveraging ConceptNet's extensive coverage of commonsense knowledge empowers the model's contextual understanding, improving the accuracy of low self-esteem detection. Human evaluation validates the effectiveness of the refined probability scores, demonstrating the model's ability to capture underlying contexts related to low self-esteem that were previously missed. Experiment results also showcase the superior performance of the proposed method compared to pre-trained language models and traditional classification models, highlighting the potential for better understanding and support for individuals experiencing low self-esteem. This study opens new possibilities for analyzing mental health expressions in social media, contributing to improved well-being and mental health interventions.

1. Introduction----------1
2. Related Work----------5
3. Methodology-----------8
4. Experiments----------29
5. Discussion-----------47
6. Conclusions----------61

[1] Stanley Coopersmith. Coopersmith self-esteem inventories. 1981.
[2] Glen Coppersmith, Mark Dredze, Craig Harman, and Kristy Hollingshead. From adhd to sad: Analyzing the language of mental health on twitter through self-reported diagnoses. In Proceedings of the 2nd workshop on computational linguistics and clinical psychology: from linguistic signal to clinical reality, pages 1–10, 2015.
[3] Munmun De Choudhury, Michael Gamon, Scott Counts, and Eric Horvitz. Predicting depression via social media. In Proceedings of the international AAAI conference on web and social media, volume 7, pages 128–137, 2013.
[4] Ana LN Fred and Anil K Jain. Learning pairwise similarity for data clustering. In 18th International Conference on Pattern Recognition (ICPR’06), volume 1, pages 925–928. IEEE, 2006.
[5] Zunaira Jamil. Monitoring tweets for depression to detect at-risk users. PhD thesis, Universit ́e d’Ottawa/University of Ottawa, 2017.
[6] Christian Karmen, Robert C Hsiung, and Thomas Wetter. Screening internet forum participants for depression symptoms by assembling and enhancing multiple nlp methods. Computer methods and programs in biomedicine, 120(1):27–36, 2015.
[7] Kurt Kroenke and Robert L Spitzer. The phq-9: a new depression diagnostic and severity measure, 2002.
[8] Baoli Li and Liping Han. Distance weighted cosine similarity measure for text classification. In Intelligent Data Engineering and Automated Learning–IDEAL 2013: 14th International Conference, IDEAL 2013, Hefei, China, October 20-23, 2013. Proceedings 14, pages 611–618. Springer, 2013.
[9] Michal Mann, Clemens MH Hosman, Herman P Schaalma, and Nanne K De Vries. Self-esteem in a broad-spectrum approach for mental health promotion. Health education research, 19(4):357–372, 2004.
[10] Keerthiram Murugesan, Mattia Atzeni, Pushkar Shukla, Mrinmaya Sachan, Pavan Kapanipathi, and Kartik Talamadupula. Enhancing text-based reinforcement learning agents with commonsense knowledge. arXiv preprint arXiv:2005.00811, 2020.
[11] Yair Neuman, Yohai Cohen, Dan Assaf, and Gabbi Kedma. Proactive screening for depression through metaphorical and automatic text analysis. Artificial intelligence in medicine, 56(1):19–25, 2012.
[12] Minsu Park, David McDonald, and Meeyoung Cha. Perception differences between the depressed and non-depressed users in twitter. In Proceedings of the international AAAI conference on web and social media, volume 7, pages 476–485, 2013.
[13] Tom Pyszczynski, Jeff Greenberg, Sheldon Solomon, Jamie Arndt, and Jeff Schimel. Why do people need self-esteem? a theoretical and empirical review. Psychological bulletin, 130(3):435, 2004.
[14] Morris Rosenberg. Rosenberg self-esteem scale (rse). Acceptance and commitment therapy. Measures package, 61(52):18, 1965.
[15] Robyn Speer, Joshua Chin, and Catherine Havasi. Conceptnet 5.5: An open multilingual graph of general knowledge. In Proceedings of the AAAI conference on artificial intelligence, volume 31, 2017.
[16] Amir Hossein Yazdavar, Hussein S Al-Olimat, Monireh Ebrahimi, Goonmeet Bajaj, Tanvi Banerjee, Krishnaprasad Thirunarayan, Jyotishman Pathak, and Amit Sheth. Semi-supervised approach to monitoring clinical depressive symptoms in social media. In Proceedings of the 2017 IEEE/ACM international conference on advances in social networks analysis and mining 2017, pages 1191–1198, 2017.
[17] Zhi-Xiu Ye, Qian Chen, Wen Wang, and Zhen-Hua Ling. Align, mask and select: A simple method for incorporating commonsense knowledge into language representation models. arXiv preprint arXiv:1908.06725, 2019.
[18] Wenpeng Yin, Jamaal Hay, and Dan Roth. Benchmarking zero-shot text classification: Datasets, evaluation and entailment approach, 2019.
[19] Pei Zhou, Karthik Gopalakrishnan, Behnam Hedayatnia, Seokhwan Kim, Jay Pujara, Xiang Ren, Yang Liu, and Dilek Hakkani-Tur. Think before you speak: Explicitly generating implicit commonsense knowledge for response generation. arXiv preprint arXiv:2110.08501, 2021.