Kensho announced this week the launch of its latest product, Kensho NERD, a cutting-edge machine learning system that unlocks the full potential of textual data by finding entities mentioned in documents, newsfeeds, or other text and linking each to a corresponding record in a database.

NERD, short for Named Entity Recognition and Disambiguation, was developed by the company’s Natural Language Processing team, which specializes in machine learning research and engineering related to analyzing and understanding human language. The first entity extraction system specifically optimized for business-related documents, it is one of several breakthrough products that the team is developing to draw insight and structure out of textual data at scale.

“We’re thrilled for the launch of NERD,” said Kensho’s CEO, Bhavesh Dayalji. “This ambitious project was the culmination of a lot of hard work and experimentation to optimize for business-related use cases. I’m looking forward to watching as NERD helps clients derive more meaning from their unstructured data and demonstrates Kensho’s transformative potential within and outside of the S&P Global ecosystem.”

NERD will see immediate internal and external use at enterprise scale, processing millions of pages of text a day in order to power smart search and augmented research workflows across its global customer base.

“NERD empowers expert analysts with supercharged research capabilities by extracting the key entities from each document they read,” said Aron Szanto, who heads the Natural Language Processing team at Kensho.

“Entity recognition is a notoriously challenging machine learning problem, but NERD’s performance approaches human accuracy, which is a reflection of the tremendous skill, creativity, and effort that went into all aspects of its development. I’m immensely proud of the team and incredibly excited to bring NERD to market.”

NERD resolves ambiguous entity mentions with AI-enabled context-awareness, handling traditionally difficult references, such as abbreviations, acronyms, aliases, and historical names.

With NERD, users can:

• Tag and analyze companies, subsidiaries and other financial organizations that appear in any textual data
• Link those entities to their entries in the S&P Global Capital IQ database to make new connections and enable deeper insights
• Connect people, places, and events to an expansive Wikimedia knowledge base containing nearly 100 million data items, providing broad analysis across domains

NERD makes it easier for users to analyze unstructured text  -  for example, investigating suppliers and competitors mentioned in company filings  -  and enables them to drill down instantaneously to the documents they need to see.

It can also be combined with other Kensho AI services, such as Kensho Link, which connects messy company data to S&P Global’s Company Database, and Kensho Scribe, a best-in-class financial audio transcription tool, to realize powerful synergies and add value to downstream analytics.

“We’ve seen many of our clients emphasize a strategic focus on unstructured data, specifically textual data, and we know NERD is a perfect tool to help derive maximum value from those sources,” said Peter Licursi, Product Manager for NLP at Kensho.

“It’s applicability to the financial services industry is undeniable, but we think NERD can provide any knowledge worker with paradigm-shifting machine learning capabilities to reinvent their research activities and content production. Our API-first approach allows any client the flexibility to implement NERD with minimal constraints and in any type of workflow.”

https://nerd.kensho.com/