We are a group of Engineers currently on Master's courses with specialisms in Machine Learning/Artificial Intelligence and have decided to apply what we are learning to help the community make decisions about unclassified data by automating the processing of this.
A general NLP system that trains and adapts using a large dynamic and publically available dataset, and then uses the classifier we have created to apply the sentiments that have been statistically calculated on keywords, phrases and quotes to summarise the sentiment behind the statement - e.g. 1-5 stars for Amazon reviews or investment predictions for Bloomberg articles. The main objective of the system is to provide a flexible and dynamic proof-of-concept of applied sentiment analysis that essentially operates on a spectrum of "positive" to "negative".
We created a front end in Java and a UI in JavaFX, which then processes small amounts of logic and data before calling the NLP system developed in python using substantial libraries and custom classifiers.
One problem we faced was that ML takes substantial time to train when running on local machines and without a hosted cloud architecture and running this code based on a 50%+ probability takes about 2-3 hours.
We are proud that we continued to work well as a team and made new friends from other universities and make links with people who also have a passion for AI and ML as we do.