Artificial Intelligence Done Right

Role of Search

It’s not really a secret that the majority of internal information is in unstructured formats. This has long been a challenge—and a frustration—for those trying to find what they need. Kamran Khan, Managing Director of Search & Content Analytics for Accenture, sees AI as the antidote to that frustration. He points out that AI is actually a “collection of multiple technologies that encompasses deep learning, natural language processing (NLP), search, ML, intelligent sensors, and robotics.” He singles out the use of NLP and ML to understand documents as a key driver to unlock the power of unstructured data.

Although NLP is not a new technology, Khan notes that it’s been going through a transformation recently. The widespread acceptance of web search engines and digital assistants pave the way for NLP applications that understand and respond to natural language queries in a similar fashion. Plus, NLP’s pairing with ML has superseded its earlier dependence on rules written out manually. That increases the scalability and accuracy of document handling.

Khan also thinks that the analysis of unstructured data is becoming more proactive and less reliant on reactive searches. It’s easier to extract relevant information from unstructured data than it was in the pre-AI, NLP, and ML days, which increases the ability of companies to “unlock the full potential of unstructured data.”

Implementing AI and Cognitive

What about AI-powered search engines, aka insight engines or cognitive search? Scott Parker, Senior Product Marketing Manager, Sinequa, has some thoughts. To do AI right, start with a valuable use case. Technology plays a role, but solving a problem takes precedence. Once you’ve got a use case, think about implementing the right methodology by acquiring data, building a model, deploying and validating your solution, and going into active learning and tuning mode. Parker stresses the importance of choosing a proven and unified technology platform so that you don’t sacrifice context or quality. Assuming you want to achieve your predicted ROI, follow his steps to move from having an AI project as an experiment to a fully implemented, successful initiative.

The dream of uncovering and optimizing the collective knowledge within an organization seems a bit closer now, thanks to AI and cognitive computing, according to Kelly Koelliker, Director of Content Marketing at Verint. The ultimate goal of AI is to get machines that can understand language as humans do. That includes disambiguating words and concepts in a contextual fashion. Understanding what is actually meant when someone says or writes something isn’t as straightforward as you might think. Humans can generally grasp the nuances—although that’s not always true—but computers have a much harder time of it. Advances in the technology, such as semantic and cognitive analysis, lead to better understanding of questions. In turn, that leads to better answers.

Koelliker also points to cognitive computing’s ability to predict what will be asked and to anticipate answers. Enhancing content automatically to incorporate implicit and explicit knowledge improvements has the potential to replace activities such as building new taxonomies or creating specialized learning. It’s exciting to contemplate how AI and cognitive computing will redefine KM.

Intelligent Content

Enhancing content to make it more intelligent is also top of mind for Mark Gross, President, Data Conversion Laboratory (DCL). For him, it’s important to realize how AI, ML, and NLP contribute to the new intelligence quotient of content and data. Calling it the “wizard behind the curtain,” DCL’s AI technologies result in data that becomes structured and curated even when it starts out as unstructured. The training sets that DCL has developed over many years amplify its ability to use technology to read the words in a document, understand the context, and structure the text so that it is no longer hidden in native “free-form” content.

Gross knows that information within organizations isn’t always easily accessible. It might even still be on paper, or electronically in Word or PDF. Embedded in these files are elements, such as tables, charts, diagrams, non-English characters, or scientific formulas. AI technologies enable DCL to turn what was previously not a cost-effective project into actionable information. Gross cites the U.S. Patent and Trademark Office (USPTO), with its millions of patent applications containing multiple formats, as a shining example of how to use AI technologies to process these applications and clear backlogs.