A multimodal approach to automatic personality recognition on Filipino social media data
There have been studies that found that linguistic style is an independent and meaningful way of exploring personality. These studies have encouraged more research on text-based personality trait recognition. With the growing trend of social media, social networking sites such as Twitter and Instagr...
Saved in:
Main Author: | |
---|---|
Format: | text |
Language: | English |
Published: |
Animo Repository
2021
|
Subjects: | |
Online Access: | https://animorepository.dlsu.edu.ph/etdm_comsci/13 https://animorepository.dlsu.edu.ph/cgi/viewcontent.cgi?article=1013&context=etdm_comsci |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | De La Salle University |
Language: | English |
Summary: | There have been studies that found that linguistic style is an independent and meaningful way of exploring personality. These studies have encouraged more research on text-based personality trait recognition. With the growing trend of social media, social networking sites such as Twitter and Instagram have gained a large amount of users, making them an easy source for data. More recent studies have looked into using different types of learning features from the sites such as texts, images, and more. Existing studies have experimented on using these features together, however, most of these studies do not experiment on low-resource language data and which features are important for the language. This research addressed this concern by making use of a filtered dataset consisting of 2231 Filipino Twitter users and experimenting on different types of features used by themselves or together. Different types of machine learning models were made for each personality trait of the Big Five with seven (7) different feature sets. Findings show that TF-IDF are the most useful features. Other text features and account features were not able to show that any learning occurred by themselves but using them together shows a small improvement in performance. Account features also show some learning with the Neuroticism personality trait. Analysis of the learning algorithms show that Support Vector Regressor (SVR) was able to perform the best but it was only able to show significant results with TF-IDF. Usage of the Radial Basis Function kernel with the SVR might have contributed to the performance.
Keywords: Natural Language Processing, Machine Learning, Psychology, Personality Traits |
---|