Project Information

Sentiment Analysis

This is a method of classifying someone’s opinion of something by analyzing a piece of text the person has written. If we gather many pieces of text about an entity written by different people, we can predict the general public’s attitude towards the entity.

Financial News

It is one of the greatest sources of current events. When people read about these current events, their feelings towards an entity may be swayed, which may also change their behaviours in the market. Eventually, the market conditions and trends will be impacted.


Current events affect human opinions and behaviours, which affect market trends. Therefore, a system that can interpret how financial news impacts people’s sentiments quickly and effectively would be useful. This project aims to develop a system that would be useful for investors to predict the market trends of an entity with relevant financial news.

Project Inspiration: FinBERT

Developed in 2019 by Araci from the University of Amsterdam, FinBERT is an NLP model designed to perform sentiment analysis on financial texts by training the already powerful NLP model BERT with large volumes of financial phrases and text.

Although FinBERT effectively finds a sentence’s overall sentiment, financial news headlines often mention multiple entities. Sometimes, these entities have different sentiments. However, FinBERT can only output one sentiment per sentence and therefore cannot distinguish the different entities and their respective sentiments. This project aims to overcome this weakness of FinBERT and develop a model that can tell separate entities in financial news headlines apart and analyze their respective sentiment.