Računalni modeli za analizu crta ličnosti i tematskih preferencija korisnika društvenih mreža

undergraduate thesis

• Title (english): Computational Models for Personality Analysis and Topic Preferences of Social Network Users
• Author: Rino Čala
• Mentor: Jan Šnajder (mentor)
• Committee member: Jan Šnajder (predsjednik povjerenstva)
• Committee member: Mladen Karan (član povjerenstva)
• Committee member: Martin Tutek (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2020-07-06, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Reddit predstavlja izrazito popularnu društvenu mrežu forumskog karaktera na kojoj pojedinci provode rasprave i dijele različite medije. Pojedini podforumi (engl. subreddits) bogati su komentarima pojedinaca i detaljima o njima. Iskoristivši ove karakteristike, formiran je skup podataka MBTI9k s komentarima i crtama ličnosti prema modelu MBTI za preko 9000 korisnika. Pomo ́cu latentne Dirichletove analize nad navedenim skupom podataka formirane su teme i korištene u predikciji ličnosti. U gradnji modela iskorištene su aktivnosti na subredditima i metoda vreće riječi. Provedena je analiza distribucije autora po formiranim temama, njihove aktivnosti na pojedinim subredditima te njihovih ponašajnih navika u pisanju komentara na objavama. Za predikciju crta ličnosti korišten je stroj s potpornim vektorima te model logističke regresije.

Abstract (english)

Reddit is a popular social network which resembles a forum-like site where users have discussions and exchange various media. Some subforums, called subreddits are rich in comments of users and their personal information. With this data a dataset called MBTI9k was formed which has comments and MBTI personalities of more than 9000 users. With latent Dirichlet analysis and using the MBTI9k dataset we formed topics from comments of users. Activities on subreddits and bag of words method were also used for machine learning models and prediction. Furthermore, we carried an analysis of the distribution of authors across formed topics, their activities across subreddits and their way of writting comments on posts. We used SVM and logistic regression models for training and prediction of personality traits.

• Keywords: Obrada prirodnog jezika; Reddit; MBTI; SVM; Logistička regresija; LDA
• Keywords (english): Natural language processing; Reddit; MBTI; SVM; Logistic regression; LDA
• Language: croatian
• URN:NBN: urn:nbn:hr:168:576863
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-13 14:43:46