This is Part 1 of a three-part series exploring the value NLP delivers across the enterprise. Parts 2 and 3 will take a deeper look at the transformative power of NLP across specific functions and through the lifecycle of a molecule.
Nearly a decade ago, the “Oracle of Silicon Valley” Tim O’Reilly famously predicted that in the information era, those with the most data win. Tim wasn’t wrong – but I’d argue that his answer was incomplete. At the time he coined the phrase, the explosion of available data was awakening organizations to the idea that information is competitive capital, and harnessing massive data repositories needed to be a core focus. And so, today what we have are enterprises who have harnessed petabytes upon petabytes of data that, for the most part, sit unused and ungoverned. What have they won?
So, I’d argue a small tweak is needed for the prediction for 2021 and beyond: Those with the most usable data win. And natural language processing or NLP, is the key to success.
Creating a tapestry of insights with NLP
Let’s look at some of the challenges enterprises face today. They are sitting on information housed in legacy systems, many of which are outdated and nearly all of which are not interconnected or able to communicate across the enterprise. Even within individual systems, information is siloed. For example, one global pharma company has clinical trial data spread across hundreds of SharePoint sites with no continuity of structure. It is costing money to maintain these systems, but abandoning them would mean losing decades of valuable data.
Then, there are acquisitions. One of our customers spent billions of dollars acquiring a biotech, largely for the intellectual property it could capture. But there was a problem – the intellectual property of interest was embedded in half a million free text documents of unknown value and unknown structure.
The result of this siloed, untapped data is akin to hundreds of individual data streams all hanging vertically like threads across the organization. No matter how many vertical threads you drape across your business, they won’t keep you warm without horizontal connectors to bring it all together.
The good news is that advances in NLP mean that we can now capture information housed across disparate systems – even the “messy” unstructured or semi-structured information – and weave it across functions for holistic enterprise-wide intelligence accessible across roles, functions and geographies. The information we unlock can automatically feed into workflows with a real-time view of the organization.
Explore, extract, empower
Those of us in the AI business tend to see the power of NLP across three key areas: exploring, extracting, and empowering. When we talk about exploring, it is the notion that a full 360-degree view of our business means pulling insights from lots of diverse sources. NLP today can ingest data from a wide range of disparate sources, including EMR extracts, call transcripts, and internal reports, all with integrated optical character recognition and table processing. Some solutions, like IQVIA’s, are purpose-built for a specific industry, so domain knowledge is embedded to ensure language is properly understood and translated. Better yet, this process can be done at an incredible speed. IQVIA’s NLP solutions can ingest more than 10 million documents per hour.
The growth rate of data is staggering. According to IDC, there were 40 zettabytes (or 44 trillion gigabytes) of data in 2019. By 2025, it is expected to grow to 175 zettabytes. Anyone who has manually parsed documents knows that documentation can include mountains of useless copy to sift and separate. NLP extraction can rapidly pull all relevant information in a way that combines and captures value points and marks up metadata so it is tagged for future use, but not bogging the user down with junk copy.
In a world where we can all download tools at will, the application of our NLP includes not only the core engine itself, but a maintained set of ontologies, as well as indexed and external content that allows us to make a meaning out of the relevant information and cut through some of the signal to noise challenges that come into play when mining text. So, for example, users can apply NLP to quickly build a complete picture of patients by automating extraction of important predictive characteristics, such as social determinants of health and lifestyle factors. In the case of our multi-billion dollar acquisition customer, NLP was able to rapidly tag and identify approximately 7,000 high value, high potential sources of content out of the half million with which it began.
Going back to our thread analogy, the only way to effectively weave insights across an enterprise is by making the outputs accessible across functions. Not everyone is an AI or NLP expert, and with today’s technology, you don’t have to be. When we talk about NLP exposing intelligence, we mean the power of today’s NLP solutions to deliver enterprise outputs in easy-to-understand visualizations. With the latest technology, NLP can transform data from different evidence streams into visual outputs to support decision-making.
Many solutions, like IQVIA’s NLP Insights Hub, have common searches pre-built into the system, making it easy to capture trends and assemble rich intelligence quickly. It is a true democratization of analytics, unlocking the ability to pull information from disparate sources to surface standardized, normalized information that can feed visualizations or other analytics tools. Those user communities who previously would have had to ask or go back to an internal analytics teams to do follow up now have easy, clickable access. To paraphrase an old car commercial, "This is not your father's NLP."
If you are in an enterprise sitting on a mountain of inanimate data, now is the time to take the next step and consider an enterprise-wide NLP solution. This is an exciting time for technology, as the long-awaited promise of rapid, enterprise-wide knowledge mining is here. No matter your industry, those who invest in a fully holistic view of their data assets and weave that knowledge effectively across their enterprise will have a valuable competitive advantage. If you’re ready to thread the needle, reach out to Linguamatics today.
For a deeper look at enterprise NLP, view our recent webinar, Step Change: Unlocking the Hidden Value of Your Enterprise’s Dark Data. You can also visit our NLP Data Factory product page to learn more about our enterprise-scale NLP solution.
Be on the lookout for Part 2 of our Enterprise NLP series next, which looks at the impact of NLP in specific roles and functions.