News

IBM's Watson AI Gets New NLP Skills for Business

• By John K. Waters
• 03/18/2020

IBM plans to enhance its Watson AI platform with several technologies from Project Debater to improve the software's ability to understand business terminology.

The enhancement will help organizations begin to identify, understand and analyze "some of the most challenging aspects of the English language" with greater clarity, the company said.

The Watson upgrade is the first commercial release of key Natural Language Processing (NLP) capabilities from Project Debater, which IBM bills as the first AI system with the ability to debate humans on complex topics. The goal of the project is to help people build persuasive arguments and make well-informed decisions. IBM introduced the project in June 2018, during the first live, public debates between AI and humans.

The project has been in development since 2012, and IBM considers it the next big milestone for AI (following the advent of Deep Blue in 1996 and Watson's performance on the TV game show "Jeopardy!" in 2011).

Project Debater was designed to digests massive texts, construct a well-structured speech on a given topic, deliver that speech with clarity and purpose, and rebut opponents' arguments. "Eventually, Project Debater will help people reason by providing compelling, evidence-based arguments, and limiting the influence of emotion, bias, or ambiguity," the company says.

"Language is a tool for expressing thought and opinion, as much as it is a tool for information," said Rob Thomas, general manager of the IBM Data and AI group, in a statement. "This is why we're harvesting technology from Project Debater and integrating it into Watson -- to enable businesses to capture, analyze, and understand more from human language and start to transform how they utilize intellectual capital that's codified in data."

The integration of the Project Debater technologies will take place throughout the coming year, the company said. The list of new capabilities includes Advanced Sentiment (analysis), Summarization (briefs) and Advanced Top Clustering.

IBM has enhanced sentiment analysis to improve Watson's ability to identify and understand complicated word schemes, such as idioms (phrases and expressions) and so-called sentiment shifters, which are combinations of words that, together, take on new meaning, such as "hardly helpful." This technology will be integrated into Watson Natural Language Understanding in March, the company said.

The company is also announcing a new classification technology designed to enable clients to create AI models that can more easily classify clauses that occur in business documents -- things like procurement contracts. With this new capability, which is based on Project Debater's deep learning-based classification tech, they can learn from as few as several hundred samples to implement new classifications quickly and easily. IBM plans to add it to its Watson Discovery enterprise search later this year.

The summarization technology pulls textual data from a variety of sources to provide users with summaries of what is being said and written about a particular topic. An early version of the Summarization capability was used analyze the red carpet "stream" at the 2020 Grammy awards. It analyzed more than 18 million articles, blogs and bios to produce bite-sized insights on hundreds of Grammy artists and celebrities.

That data was then "infused" into the red carpet live stream, on-demand videos and photos across www.grammy.com "to give fans deeper context about the leading topics of the night." IBM plans to add this capability to IBM Watson Natural Language Understanding later in the year.

New topic clustering techniques built on insights gained from Project Debater will enable users to "cluster" incoming data to create meaningful "topics" of related information, which can then be analyzed. The technique, which IBM plans to integrate into Watson Discovery later this year, will also allow subject-matter experts to customize and fine-tune the topics to reflect the language of specific businesses or industries, such as insurance, health care and manufacturing.