Abstract:
Sentiment analysis as a sub- eld of natural language processing has received increased
attention in the past decade enabling organisations to more e ectively manage their
reputation through online media monitoring. Many drivers impact reputation, however,
this thesis focuses only the aspect of nancial performance and explores the gap with
regards to nancial sentiment analysis in a South African context.
Results showed that pre-trained sentiment analysers are least e ective for this task
and that traditional lexicon-based and machine learning approaches are best suited to
predict nancial sentiment of news articles. The study contributed to updating an existing
sentiment dictionary and developing a full pipeline to lter data for nancial topics
and predict sentiment. Using a binary logistic regression model and a binary XGBoost
classi er on both headlines and article content produced accuracies of >85%. The predicted
sentiments correlated quite well with share price and highlighted the potential
use of sentiment as an indicator of nancial performance.
Model generalisation was less acceptable due to the limited amount of training data
used. Future work includes expanding the data set to improve general usability and
contribute to an open-source nancial sentiment analyser for South African data.
