Basit öğe kaydını göster

dc.contributor.advisorDehkharghani, Rahimen_US
dc.contributor.authorArmah, Courageen_US
dc.contributor.otherIşık Üniversitesi, Lisansüstü Eğitim Enstitüsü, Bilgisayar Mühendisliği Yüksek Lisans Programıen_US
dc.contributor.otherIşık University, School of Graduate Studies, Computer Science Engineering Master Programen_US
dc.date.accessioned2024-03-14T16:40:28Z
dc.date.available2024-03-14T16:40:28Z
dc.date.issued2024-02-12
dc.identifier.citationArmah, C. (2024). Multi-task learning on mental disorder detection, sentiment detection and emotion detection. İstanbul: Işık Üniversitesi Lisansüstü Eğitim Enstitüsü.en_US
dc.identifier.urihttps://hdl.handle.net/11729/5919
dc.descriptionText in English ; Abstract: English and Turkishen_US
dc.descriptionIncludes bibliographical references (leaves 41-45)en_US
dc.descriptionx, 49 leavesen_US
dc.description.abstractSuicidal behavior is a global cause of life-threatening injury and most of the time, death. Mental disorders such as depression, anxiety, and bipolar are prevalent among the youth in recent decades. Social media are popular platforms for individuals to post their thoughts and feelings on. Extracting people’s sentiments and feelings from such online platforms would help detect mental disorders of the users to treat them before it becomes too late. This thesis investigates the use of multi-task learning systems and single-task learning techniques to estimate behaviors and mental states for early diagnosis. I used data mined from Reddit, one of the popular social media platforms that provides anonymity. Anonymity increases the chances of individuals sharing what they truly feel in their real life. The obtained results by the proposed approaches open new doors to the understanding of how multi-task systems can increase the performance of text classification problems such as depression detection, emotion detection, and sentiment analysis, trained together in a multi-task learning network when compared to their training in isolation in a single-task learning network. We used the SWMH dataset, already labeled by 5 different depression labels (depression, anxiety, suicide, bipolar, and off my chest) and then added emotion and polarity labels to it and made it publicly available for researchers in the literature. The obtained results in this study are also comparable to other approaches in the field.en_US
dc.description.abstractİntihar düşüncesi, dünya çapında, ömür boyu tehdit eden yaralanmaların ve çoğu zaman ölümün bir nedenidir. Depresyon, ankseyete bozukluğu ve bipolar gibi zihinsel bozukluklar, son yıllarda gençler arasında yaygındır. Sosyal medya, bireylerin duygu ve düşüncelerini paylaştıkları popüler platformlardır. Sosyal medya platformlardan insanların duygu ve düşüncelerinin çıkarılması, uzmanlar kullanıcıların zihinsel bozukluklarınını tespit edilmesine ve çok geç olmadan tedavi edilmesine yardımcı olacaktır. Bu tez, erken tanı için davranışları ve zihinsel durumları tahmin etmeye yönelik, çok görevli öğrenme sistemlerinin ve derin öğrenme tekniklerinin kullanımını araştırmaya çalışmaktadır. Anonimlik sağlayan popüler sosyal medya platformlarından biri olan Reddit'in metin verilerini kullandım. Anonimlik, bireylerin gerçek yaşamlarında hissettiklerini paylaşmasına artırır. Önerilen yaklaşımlarla elde edilen sonuçlar, çok görevli sistemlerin, izole eğitimlerine kıyasla birlikte eğitilen depresyon tespiti, duygu tespiti ve duygu analizi gibi metin sınıflandırma problemlerinin performansını nasıl artırabileceğinin anlaşılmasına yeni kapılar açmaktadır. Bu çalışmada elde edilen sonuçlar, alandaki diğer yaklaşımlarla da karşılaştırılabilir niteliktedir.en_US
dc.description.tableofcontentsPurpose of Studyen_US
dc.description.tableofcontentsLITERATURE REVIEWen_US
dc.description.tableofcontentsTheoretical Backgrounden_US
dc.description.tableofcontentsRelated Worksen_US
dc.description.tableofcontentsSUGGESTED APPROACHen_US
dc.description.tableofcontentsData Pre-Processingen_US
dc.description.tableofcontentsGeneral Methodologyen_US
dc.description.tableofcontentsEXPERIMENTAL EVALUATIONen_US
dc.description.tableofcontentsDataseten_US
dc.description.tableofcontentsUsed Programming Languages, Platforms and Toolsen_US
dc.description.tableofcontentsMetrics for Evaluationen_US
dc.description.tableofcontentsSingle Task with Attention Layer vs MTL with Attention Layeren_US
dc.description.tableofcontentsSingle Task without Attention Layer vs MTL with no Attention Layeren_US
dc.description.tableofcontentsA subset of original dataseten_US
dc.description.tableofcontentsIllustrative table that shows basic elements to explain evaluation metricsen_US
dc.description.tableofcontentsExperimental results of f1-score from test dataset compared with other research in the fielden_US
dc.description.tableofcontentsSample predictions of custom single task and proposed MTL modelsen_US
dc.description.tableofcontentsMultitask learning frameworks according to relatednessen_US
dc.description.tableofcontentsThe General view of the suggested Methodologyen_US
dc.description.tableofcontentsCustom single-task model with attention layeren_US
dc.description.tableofcontentsCustom single-task model without attention layeren_US
dc.description.tableofcontentsThe proposed MTL modelen_US
dc.description.tableofcontentsSubset of Extended SWMH dataset after labelling by polarity and emotion labelsen_US
dc.description.tableofcontentsClassification report of mental disorder detection using custom model that has an attention layeren_US
dc.description.tableofcontentsClassification report of sentiment detection using a custom model that has an attention layeren_US
dc.description.tableofcontentsClassification report of emotion detection task using a custom model that has an attention layeren_US
dc.description.tableofcontentsClassification report of MTL using a custom model with attention layeren_US
dc.description.tableofcontentsClassification report of mental disorder detection using the custom model without an attention layeren_US
dc.description.tableofcontentsClassification report of sentiment detection using the custom model without an attention layeren_US
dc.description.tableofcontentsClassification report of emotion detection using the custom model without an attention layeren_US
dc.description.tableofcontentsMTL classification report using the custom model without an attention layeren_US
dc.description.tableofcontentsCorrelation matrix showing the relation between mental disorder labels, sentiment and emotion labelsen_US
dc.description.tableofcontentsConfusion matrix showing mental disorder evaluations for both single task and MTL models respectivelyen_US
dc.description.tableofcontentsConfusion matrix showing sentiment detection evaluations for both single task and MTL models respectivelyen_US
dc.language.isoenen_US
dc.publisherIşık Üniversitesien_US
dc.rightsinfo:eu-repo/semantics/openAccessen_US
dc.rightsAttribution-NonCommercial-NoDerivs 3.0 United States*
dc.rights.urihttp://creativecommons.org/licenses/by-nc-nd/3.0/us/*
dc.subjectNatural language processingen_US
dc.subjectMulti-task learningen_US
dc.subjectSentiment analysisen_US
dc.subjectDeep learningen_US
dc.subjectEmotion detectionen_US
dc.subjectDoğal dil işlemeen_US
dc.subjectDuygu analizien_US
dc.subjectDuygu algılamaen_US
dc.subjectDerin öğrenmeen_US
dc.subjectÇok görevli öğrenmeen_US
dc.titleMulti-task learning on mental disorder detection, sentiment detection and emotion detectionen_US
dc.title.alternativeZihinsel bozukluk tespiti, duygusallık(sentiment) tespiti ve duygu tespiti üzerinde çok görevli öğrenimen_US
dc.typeMaster Thesisen_US
dc.departmentIşık Üniversitesi, Lisansüstü Eğitim Enstitüsü, Bilgisayar Mühendisliği Yüksek Lisans Programıen_US
dc.departmentIşık University, School of Graduate Studies, Computer Science Engineering Master Programen_US
dc.authorid0000-0001-5765-5735
dc.authorid0000-0001-5765-5735en_US
dc.relation.publicationcategoryTezen_US
dc.institutionauthorArmah, Courageen_US


Bu öğenin dosyaları:

Thumbnail

Bu öğe aşağıdaki koleksiyon(lar)da görünmektedir.

Basit öğe kaydını göster

info:eu-repo/semantics/openAccess
Aksi belirtilmediği sürece bu öğenin lisansı: info:eu-repo/semantics/openAccess