11
Kensho: A Hub for AI Innovation and Transformation

Since becoming part of S&P Global in 2018, Kensho has been building and deploying transformational technology solutions for S&P Global as well as for external clients. By applying its AI and machine-learning capabilities, Kensho is accelerating and automating core workflows and transforming the way data is ingested, linked and tagged, and how insights are discovered. Kensho’s core capabilities include machine learning, natural-language processing (NLP), and data discovery. S&P Global is combining Kensho’s AI/ML technology with its unparalleled datasets to create a significant competitive advantage.

Kensho is combining advanced technology and a start-up mindset to create new revenue streams, improve time-to-market, and reduce costs across S&P Global’s divisions.

Machine-Learning


Applying AI solutions that structure data and derive insights in a timely manner at scale

Example:
Kensho Scribe transcribes financial and business audio into text with unparalleled accuracy and speed. The only speech-to-text product custom built for finance, it handles complexities including jargon, accents, numbers, currencies, and product names as well as the nuances of spoken language, including mumbling, stuttering, and self-correction.
Case Study:
Scribe is being used by a leading expert network service, which provides its customers a searchable database of thousands of transcripts of conversations between buy-side investors and industry experts. Accuracy and compliance are critical to the customer, and they use 25,000 hours annually of Scribe’s highly accurate transcription service and layer on NERD capabilities to highlight client names and other critical data and accelerate their compliance workflow.

Natural-Language Processing


Transforming business processes through innovation  

Example:
Kensho NERD is a cutting-edge machine-learning system that unlocks the full potential of text documents by identifying the people, places, organizations and events in that data and linking them to deep sources of structured knowledge including the Capital IQ database. Trained on vast quantities of S&P’s proprietary data, NERD has emerged as the world’s foremost entity extraction product for business, processing trillions of characters of text to power Codex, MI data feeds, and external customer workflows.
Case Study:
NERD on Xpressfeed: Machine Readable Transcripts. The integration of NERD into the Transcripts dataset provides MI’s clients with access to 60 million new links to Capital IQ’s company database, across more than a million machine readable transcripts and counting, enabling new insights and downstream analytics.

Data Visualization


Make data discoverable

Example:
Codex is an AI-powered document-viewing platform that facilitates the efficient discovery and extraction of relevant information from large volumes of text—enabling timely, fact-based decision-making. It includes smart natural-language search with synonym support, search across all company documents, keyword matches within tables, and other features for quickly locating essential information.
Case Study:
Codex DocViewer went live as part of S&P Capital IQ Pro in September 2021. Since the launch, hundreds of thousands of documents have been viewed using Codex, and tens of thousands of document searches have been conducted, providing next-gen interaction features to S&P Capital Pro users.

Incubate early-stage ideas for white space products and services  

Example:
Kensho Classify is a breakthrough AI capability that attains human-level reading comprehension at machine scale by understanding the key topics, concepts, and themes relevant to any text document. Classify’s self-teaching capabilities enable customers across industries to seamlessly customize their model’s domain of expertise, powering deep analytics and intelligent search workflows.
Example:
SPGISpeech is a large-scale transcription dataset developed by Kensho’s R&D team that leverages S&P Global’s rich data resources. With 5,000 hours of fully formatted, professionally transcribed audio, it contains a broad cross section of accents, strongly varying audio quality, and both spontaneous and narrated speech. The subject of several research and academic papers, SPGISpeech is being used by academic institutions around the world to help address the data gap between industry and academic institutions.