A Deep Learning Approach for Automated Depression Assessment Using Roman Urdu

Depression is a highly complex and frequently unnoticed mental condition that presents substantial hazards to a person's well-being. Many people use social media as a means of conveying their emotions in the digital age, creating a potential avenue for depression assessment through these platfo...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:IEEE access Ročník 12; s. 193387 - 193401
Hlavní autoři: Mohmand, Ruba, Habib, Usman, Usman, Muhammad, Baili, Jamel, Nam, Yunyoung
Médium: Journal Article
Jazyk:angličtina
Vydáno: Piscataway IEEE 2024
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Témata:
Accuracy
Deep learning
Depression
Feature extraction
Hate speech
Hazard assessment
Hazard identification
health care
Mental depression
Mental health
Natural language processing
Performance evaluation
Sentiment analysis
Social networking (online)
text classification
Transfer learning
Transformers
ISSN:2169-3536, 2169-3536
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Depression is a highly complex and frequently unnoticed mental condition that presents substantial hazards to a person's well-being. Many people use social media as a means of conveying their emotions in the digital age, creating a potential avenue for depression assessment through these platforms. While English has seen extensive research on depression identification, other languages, like Roman Urdu, which is widely used in South Asia, have received less investigation. The lack of a standardized textual structure in Roman Urdu presents a challenge in evaluating depression levels. Moreover, there is currently no corpus available to assess the severity of depression in Roman Urdu. This is a notable deficiency considering the significance of linguistic resources for tasks involving natural language processing. This study fills this gap by manually categorizing an extensive collection of 25,004 Roman Urdu posts from X(Twitter) into four classes: mild, moderate, severe, and non-depression. Using a pre-trained BERT model with transfer learning, the study achieves a remarkable accuracy rate of 99%, surpassing the performance of other deep models. The findings underscore the capacity of sophisticated natural language processing methods to precisely evaluate the intensity of depression in Roman Urdu text, thereby facilitating more comprehensive mental health assessments.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:2169-3536
2169-3536
DOI:10.1109/ACCESS.2024.3519264