In April we had the chance to collaborate with Bloomberg to offer you a sneak peak into some machine learning that they do on a daily basis. In case you missed it, you can view the video here. Thank you to Bloomberg for sponsoring the event, space, and refreshments!

The event had four speakers:

Sentiment Analysis : From Bag of Words to Bag of Insights – Slides

Vika Abrecht talked about the evolution of sentiment analysis projects at Bloomberg. She showed how sentiment analysis concepts and techniques can be applied to various sources and a range of targets to extract valuable financial insights.

News Story Classification by Feature Extraction and Ranking – Slides

Anna Cianciara presented a novel alternative to statistical classification of news stories: a method to enhance classification rules in Bloomberg’s existing proprietary framework by extracting features from unstructured text and ranking key terms and phrases according to their predictive value.

How to Train Your Search Engine – Slides

Haoyun Feng talked about how Bloomberg applies machine learning algorithms to build a search engine that helps their clients discover and navigate information inside the Bloomberg terminal which aggregates, stores, and classifies more than one million news stories from over 100,000 sources daily and brings together 350+ exchanges, 4,000+ Foreign Exchange feeds and 80,000+ newswires from across the world.

Practical Considerations for ML Evaluation at Scale – Slides

Leslie Barrett covered some of the most useful evaluation metrics for doing machine learning evaluation at scale, in a dynamic enterprise environment, where crucial product decisions must be made on the basis of this data.