Bloomberg

OptimalAI scientists supplement Bloomberg's AI team to apply NLP, ML and deep learning to the Terminal

Bloomberg Terminal

The Bloomberg Terminal is the world's largest software system for financial service professionals, providing access to more than 35 million financial instruments across all asset classes.

In applying AI to the Terminals, Bloomberg employs over 200 data scientists, including former professors and graduates from internationally-renowned programs. For over a decade, OptimalAI scientists have supplemented Bloomberg's internal team to help apply natural language processing, machine learning and deep learning to core document understanding, recommendations and the ongoing evolution of the platform.

Bloomberg: Understanding Bloomberg and The Terminal (2m 13s)

NLP

The Bloomberg Terminal utilises a real-time NLP library to perform low-level text resolution tasks, such as tokenization, chunking and parsing. Sitting on top of this, named entity extractors detect people, companies, tickers and organizations found in the text of news and social text databases. These named entity extractors enable sentiment analysis functions that estimate how positive or negative a piece of news is for a particular company.

Information Extraction

Of the broad number of AI/ML applications, computer vision and/or NLP algorithms are used to extract semantic meaning and relationships from video, audio, blog posts, tweets, and more. The large suite of tools for structured and unstructured data, include table detection and segmentation tools that enable analysts to increase their scope of ingested data, as well as systems for figure understanding that extract the underlying data from scatter plots.

There is also significant functionality that connects text to other artifacts, such as people or stock tickers. News importance indicators on the Terminal automatically detect and tag crucial headlines, while a robust related stories function highlights relevant additional information to readers.

Search

In addition to a sophisticated search system that ranks and queries understanding, a natural language query interface enables Terminal users to ask questions in plain English and receive precise answers. This search functionality is deployed across document collections, with a particular focus on news search and ranking function. The internal help system uses automatic routing systems to direct queries to the right experts, while automatic answering capabilities detect and answer frequently-recurring customer inquiries.

Quality Control

For quality control, anomaly detection algorithms serve as critical tools for identifying inconsistencies in datasets, thereby ensuring data precision. These methods find hidden investment opportunities and flag suspicious market activity.

For instance, should a financial analyst alter the evaluation of a stock post the company’s quarterly earnings disclosure, the anomaly detection will determine whether such an action falls within a typical behavioral paradigm or signify an exceptional occurrence. If the latter, this information could be pertinent for Bloomberg clients, potentially influencing their investment decisions.

Insight Generation

For insight generation, AI/ML systems analyze large datasets to unlock investment signals that might not otherwise be observed. For example, using highly correlated data like credit card transactions to gain visibility into recent company performance and consumer trends. Or, analyzing millions of daily news articles to understand the questions and themes influencing markets, economic sectors and trading volumes in specific corporate securities.