基于层次数据增强的多维度特征融合社交媒体抑郁症识别

doi:10.3778/j.issn.1002-8331.2406-0368

摘要/Abstract

摘要： 抑郁症作为一种常见的精神障碍，因其数据样本稀缺且难以获取，抑郁识别任务在特征的捕获方面正面临极大挑战。提出了基于层次的数据增强技术，通过同义词替换和句法树结构调整方法生成新的抑郁文本，丰富抑郁数据集。在此基础上，构建了一个融合多维度特征的社交媒体抑郁症识别模型，该模型整合了文本的风格特征、情感特征和语境特征，并引入多头注意力机制，突出抑郁特征中的关键信息，从而对抑郁文本进行精准分类。实验结果表明，提出的方法有效扩充了样本数据，并在多维度上准确提取了抑郁特征，使得抑郁识别的准确率达到了92%，验证了模型的有效性。

关键词: 社交媒体, 抑郁症识别, 数据增强, 多维度特征

Abstract: Depression, a common mental disorder, poses significant challenges in feature extraction for recognition due to the scarcity of data samples. This study proposes a hierarchical data augmentation technique, generating new depressive text samples through synonym replacement and syntactic tree adjustment, enriching the dataset. A multidimensional feature fusion model for social media depression recognition is then constructed, integrating stylistic, emotional, and contextual features, with a multi-head attention mechanism emphasizing key depressive information for precise text classification. Experimental results show that this method effectively expands the sample data and accurately extracts depressive features, achieving a recognition accuracy of 92%, confirming the model’s effectiveness.

Key words: social media, depression recognition, data augmentation, multi-dimensional feature

李世琪, 刁宇峰, 张浩, 杨亮, 林鸿飞, 樊小超. 基于层次数据增强的多维度特征融合社交媒体抑郁症识别[J]. 计算机工程与应用, 2025, 61(19): 127-136.

LI Shiqi, DIAO Yufeng, ZHANG Hao, YANG Liang, LIN Hongfei, FAN Xiaochao. Hierarchical Data Augmentation Based Multi-Dimensional Feature Fusion for Social Media Depression Recognition[J]. Computer Engineering and Applications, 2025, 61(19): 127-136.

参考文献

[1] 杨权. 抑郁障碍的诊断与治疗[M]. 成都: 四川科学技术出版社, 2003.
YANG Q. Diagnosis and treatment of depressive disorder[M]. Chengdu: Sichuan Scientific & Technical Publishers, 2003.
[2] FRIEDMAN S. Depression: clinical, experimental, and theoretical aspects[J]. The Journal of the American Medical Association, 1968, 203(13): 1144.
[3] RAMIREZ-ESPARZA N, CHUNG C, KACEWIC E, et al. The psychology of word use in depression forums in English and in Spanish: testing two text analytic approaches[J]. Proceedings of the International AAAI Conference on Web and Social Media, 2008, 2(1): 102-108.
[4] SPITZER R L, WILLIAMS J B, GIBBON M, et al. The structured clinical interview for DSM-III-R (SCID). I: history, rationale, and description[J]. Archives of General Psychiatry, 1992, 49(8): 624-629.
[5] SHEEHAN D V, LECRUBIER Y, SHEEHAN K H, et al. The mini-international neuropsychiatric interview (M.I.N.I.): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10[J]. J Clin Psychiatry, 1998, 59(S20): 22-33.
[6] MONTGOMERY S A, ASBERG M. A new depression scale designed to be sensitive to change[J]. British Journal of Psychiatry, 1979, 134: 382-389.
[7] BECK A T, STEER R A, CARBIN M G. Psychometric properties of the beck depression inventory: twenty-five years of evaluation[J]. Clinical Psychology Review, 1988, 8(1): 77-100.
[8] KROENKE K, SPITZER R L. The PHQ-9: a new depression diagnostic and severity measure[J]. Psychiatric Annals, 2002, 32(9): 509-515.
[9] MUKHIYA S K, AHMED U, RABBI F, et al. Adaptation of IDPT system based on patient-authored text data using NLP[C]//Proceedings of the 2020 IEEE 33rd International Symposium on Computer-Based Medical Systems. Piscataway: IEEE, 2020: 226-232.
[10] REECE A G, REAGAN A J, LIX K L M, et al. Forecasting the onset and course of mental illness with Twitter data[J]. Scientific Reports, 2017, 7(1): 13006.
[11] 周莹. 多类型特征融合的抑郁倾向识别方法研究[D]. 济南: 山东师范大学, 2019.
ZHOU Y. Research on the identification method of depressive tendency based on multi-type feature fusion[D]. Jinan: Shandong Normal University, 2019.
[12] SEAL A, BAJPAI R, AGNIHOTRI J, et al. DeprNet: a deep convolution neural network framework for detecting depression using EEG[J]. IEEE Transactions on Instrumentation and Measurement, 2021, 70: 2505413.
[13] YANG L J, WANG Y X, ZHU X R, et al. A gated temporal-separable attention network for EEG-based depression recognition[J]. Computers in Biology and Medicine, 2023, 157: 106782.
[14] TRIFAN A, ANTUNES R, MATOS S, et al. Understanding depression from psycholinguistic patterns in social media texts[C]//Advances in Information Retrieval. Cham: Springer, 2020: 402-409.
[15] ZOGAN H, RAZZAK I, WANG X Z, et al. Explainable depression detection with multi-aspect features using a hybrid deep learning model on social media[J]. World Wide Web, 2022, 25(1): 281-304.
[16] BELINKOV Y, BISK Y. Synthetic and natural noise both break neural machine translation[J]. arXiv:1711.02173, 2017.
[17] ZHANG X, ZHAO J, LECUN Y. Character-level convolutional networks for text classification[C]//Advances in Neural Information Processing Systems, 2015.
[18] WU X, LV S W, ZANG L J, et al. Conditional BERT contextual augmentation[C]//Proceedings of the 19th International Conference on Computational Science. Cham: Springer, 2019: 84-95.
[19] WEI J, ZOU K. EDA: easy data augmentation techniques for boosting performance on text classification tasks[J]. arXiv:1901.11196, 2019.
[20] BROWN T, MANN B, RYDER N, et al. Language models are few-shot learners[C]//Advances in Neural Information Processing Systems, 2020: 1877-1901.
[21] DAI H, LIU Z, LIAO W, et al. AugGPT: leveraging ChatGPT for text data augmentation [J]. arXiv:2302.13007, 2023.
[22] PENNEBAKER J W, MEHL M R, NIEDERHOFFER K G. Psychological aspects of natural language use: our words, our selves[J]. Annual Review of Psychology, 2003, 54: 547-577.
[23] STIRMAN S W, PENNEBAKER J W. Word use in the poetry of suicidal and nonsuicidal poets[J]. Psychosomatic Medicine, 2001, 63(4): 517-522.
[24] PO?WIATA R, PERE?KIEWICZ M. OPI@LT-EDI-ACL2022: detecting signs of depression from social media text using RoBERTa pre-trained language models[C]//Proceedings of the Second Workshop on Language Technology for Equality, Diversity and Inclusion. Stroudsburg: ACL, 2022: 276-282.
[25] JIANG B, ZHANG Z Y, LIN D D, et al. Semi-supervised learning with graph learning-convolutional networks[C]//Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2019: 11305-11312.
[26] SHEN Y, TAN S, SORDONI A, et al. Ordered neurons: integrating tree structures into recurrent neural networks[J]. arXiv:1810.09536, 2018.
[27] CORTES C, VAPNIK V. Support-vector networks[J]. Machine Learning, 1995, 20(3): 273-297.
[28] GROSSBERG S. Recurrent neural networks[J]. Scholarpedia, 2013, 8(2): 1888.
[29] LAN Z, CHEN M, GOODMAN S, et al. ALBERT: a lite BERT for self-supervised learning of language representations[J]. arXiv:1909.11942, 2019.
[30] CHEN Y. Convolutional neural network for sentence classification[D]. Waterloo: University of Waterloo, 2015.
[31] BREIMAN L. Random forests[J]. Machine Learning, 2001, 45(1): 5-32.
[32] LECUN Y, BENGIO Y. Convolutional networks for images, speech, and time series[M]//The handbook of brain theory and neural networks. Cambridge: MIT Press, 1995.
[33] DEVLIN J, CHANG M W, LEE K, et al. BERT: pre-training of deep bidirectional transformers for language understanding[J]. arXiv:1810.04805, 2018.
[34] JI S, ZHANG T, ANSARI L, et al. MentalBERT: publicly available pretrained language models for mental healthcare[J]. arXiv:2110.15621, 2021.