| Summary: | Social Media has been thoroughly integrated into the many facets of societies across the world,
churning out vast quantities of valuable data that hides a multitude of insights. In recent years,
many novel techniques and methods have been brought to light and made mainstream through
open-source repositories. These cutting-edge tools have allowed applicants of the technology to
rapidly produce a multitude of applications that extract insights from social media data.
Attempted here will be a social media data mining pipeline to perform automated personality
assessment and evaluation. This pipeline consists of 5 stages in sequence; data collection, data
transformation, data preprocessing, model execution and personality evaluation. To discover how
best to implement each stage, exploratory analysis and experiments were conducted for
familiarising with the materials and comparison’s sake respectively.
Primary to the pipeline is an analysis and classification on social media users’ personalities through
analysing their historical timeline laced with their opinions, comments, ideas, and interactions.
Each tweet will be analysed for its sentiment, emotion, and personality traits. Consulting the big-five personality trait model, behavioural classification using pre-built models, transformers and
Zero-Shot classification will be used. Additionally, the pipeline will be tested by feeding thousands
of tweets collected from Twitter using API scraping methods. The pipeline was then later deployed
onto a web application as a proof of concept (PoC) implemented using Streamlit which also
includes various visualisations and options for customising the pipeline.
|
|---|