AI learns the language of business: Behind the scenes at IBM’s NLP Innovation Event
At IBM’s inaugural NLP innovation event, held at IBM Watson’s Astor Place Headquarters last week, Rob Thomas, General Manager, IBM Data and AI, announced that Watson’s core AI products will include exciting enhanced Natural Language Processing (NLP) capabilities. The new technologies represent the first commercialization of key NLP capabilities to come from IBM Research’s Project Debater. With these new capabilities Watson will better understand the way humans communicate with each other, by weighing multiple perspectives, eliminating irrelevant data, and providing tools to help users more easily retrieve and gain insights from language-based data.
The key new features IBM Watson announced are:
- Advanced Sentiment Analysis: Watson can better identify and understand complicated, sometimes colloquial word schemes like idioms and sentiment shifters, such as “hardly helpful,” and their meanings. This feature also includes new classification technology that will enable clients to create NLP models that will more easily classify industry-specific business language.
- Summarization: Watson can analyze a variety of sources and provide a summary or brief of the ideas and information contained within. An early iteration of this technology was used at the 2020 Grammy Awards during the red-carpet livestream.
- Clustering: New topic clustering techniques give subject matter experts a tool to cluster incoming data into topics so they can peruse topics and view buckets of related information. They can then fine-tune topics to reflect the language of specific businesses or industries, like insurance, healthcare, and manufacturing.
This marks the first commercial release of key NLP capabilities from IBM Research’s Project Debater, IBM’s breakthrough AI system built to debate humans on complex topics. The goal of Project Debater is to help people build persuasive arguments and make well informed decisions regardless of their personal bias.
Following Thomas’ introduction, he invited a host of NLP experts to the stage for panels and talks organized around IBM’s guiding principles for building market-leading NLP, which we call the NLP framework. The NLP framework consists of four NLP tasks that are integral to teaching Watson how to fully engage with human languages.
Ritika Gunnar, Vice President, Data & AI Expert Labs & Learning, then took our expertise down a deep dive into the nuances of NLP, the NLP framework, and client use cases. The panelists were Aya Soffer, Vice President AI Technologies, IBM Research (calling in from the research lab in Haifa Israel); Ruchir Puri, Chief Scientist, IBM Research; and myself. In this panel we covered the history of the field of NLP, how NLP fits into the larger AI universe, and companies that are using Watson applications to successfully apply natural language processing.
One capability powered by IBM’s advanced NLP is Watson Discovery’s Smart Document Understanding feature (SDU). Using SDU, a large automotive company has been able to train Watson to understand the contents of manuals, and provided car owners with a simple chatbot they can use to ask questions and quickly surface answers. Information in car manuals is organized in different sections and tables, each of which has its own meaning, and now Watson is uniquely suited to analyze and retrieve just the information needed from the entirety of a manual’s contents.
In another example, a professional services company is using Watson to help users get the maximum R&D tax credits allowable for an enterprise. No two enterprises engaged in R&D are the same, so in this instance Watson needs to understand which tax laws a specific enterprise will benefit from.
If all of this sounds intriguing, no worries, you didn’t miss a thing. To check out these panels, announcements and demos, tune into our digital simulive event and see it all for yourself.