Tech At Bloomberg

Artificial Intelligence (AI)

Artificial intelligence (AI) disciplines, such as natural language processing (NLP) and machine learning (ML), play a central role at Bloomberg.

We have always relied on text as a key underlying source of data for our clients. Over the past decade, we have made significant investments in statistical NLP techniques, extending our capabilities by building state-of-the-art technology for core document understanding, recommendation and customer-facing systems.

The vital role of NLP at Bloomberg

Our NLP technology extracts structured information from documents in a process known as digitization or normalization. At the core of this program is a proprietary, robust real-time NLP library that performs low-level text resolution tasks, such as tokenization, chunking and parsing. Sitting on top of this core toolset, named entity extractors detect people, companies, tickers and organizations found in the natural text in our news and social text databases.

These named entity extractors are crucial for enabling our sentiment analysis functions on the Bloomberg Terminal (BSV <GO> and TREN <GO>), derived indicators that estimate how positive or negative a piece of news is for a particular company. Beyond that, our topic classification engine automatically tags documents with normalized topics (such as oil) to make retrieval and monitoring straightforward. In the legal domain, we have built a legal principles engine that enables lawyers to uncover the underlying case law argumentation that supports a particular decision.

Extracting meaningful insights

Beyond these core functions, we have developed sophisticated fact extractors (or relationship extractors) that pick out specific information from documents in order to ease our ingestion flow. We have also built a large suite of tools for structured data, including our table detection and segmentation tools that enable our analysts to increase their scope of ingested data, as well as systems for figure understanding that extract the underlying data from scatter plots. Our self-service topic streams enable our reporters to find news about the companies or sectors they cover.

While these core NLP tools stay strictly within the domain of text, we have built significant functionality that connects text to other artifacts, such as people or stock tickers. Our news importance indicators on the Terminal automatically detect and tag crucial headlines, while a robust related stories function highlights relevant additional information to readers.

Helping customers find what they need, faster

In addition to our sophisticated search system (HL <GO>), which features state-of-the-art ranking and query understanding, our natural language query interface enables Terminal users to ask questions in plain English – and get precise answers. This search functionality is deployed across many document collections, with a particular focus on our news search and ranking function (NSE <GO>). Our internal help system uses automatic routing systems to direct queries to the right experts, while our automatic answering capabilities detect and answer frequently-recurring customer inquiries.

A growing team committed to the community

Bloomberg employs over 200 NLP and ML experts, including former professors and graduates from internationally-renowned programs. As the team expands, so does the infrastructure we build to support them – such as the large GPU cluster we use to speed up the deep learning/neural network models that make up an increasingly large part of our deployed technology.

Our published AI research

Bloomberg contributes back to academia whenever we can by attending and speaking at conferences in ML, NLP, and IR, handing out the Bloomberg Data Science Research Grant, hosting the Bloomberg Data Science Ph.D. Fellows and serving as committee members for conferences.