-
Landing Page
-
Text Classification
-
Classification Result
Inspiration
The idea behind the project came from an interest in natural language processing (NLP) and financial markets, as well as a desire to reduce time spent reading business news
What it does
The project sets out to predict the sentiment in financial news reports to gauge analysts' perceptions about a publicly traded company.
How I built it
Most experiments were conducted on a Naive Bayes classifier model due to its simplicity of application, performance in text classification, and low computational cost.
Technologies used: Python, Flask, HTML, CSS, Spacy NLP library, Scikit-learn library.
Challenges
The primary challenges revolved around imbalanced datasets and text preprocessing.
Accomplishments
75%-80% accuracy
What I learned
End-to-end NLP pipeline + implementation of model on Webb app
What's next
The main improvements to the project would arise from implementing a web scrapper to allow users to copy-paste a URL and automatically gather the article's content, in addition to experimenting with deep neural networks (transformers) for increased accuracy.
Log in or sign up for Devpost to join the conversation.