How Dell Technologies and NVIDIA Support Natural Language Processing Technologies

I previously talked about the Rise of Deep Learning in the Enterprise and how its use is dramatically augmenting human capabilities. Gartner is predicting that the Artificial Intelligence (AI) Augmentation market will be $2.9 trillion by 2021. Let’s think about that number. Globally, there are only three countries with a GDP larger than $2.9 trillion, which means the AI market will be larger than most countries’ total GDP. Just a hint here: if you struggle to get an IT project green lighted, try incorporating an AI initiative into it. The value of AI is one of the primary reasons why enterprises are fast-tracking AI infrastructure projects. In this new blog series, we’ll focus on specific types of Deep Learning (DL) use cases and their impacts in the enterprise. The first one is Natural Language Processing (NLP).

Natural Language Processing is the original end state dream of AI researchers. In fact, a key basis of the Turing Test is determining the ability of a machine to understand human language and respond. It was created by Alan Turing in 1950 to determine whether a computer can think. The basic tenet is to test if a machine can use language to fool a human into thinking it is a human being. Even 70 years later we have yet to convincingly pass the Turing test because, to put it simply, human language is hard.

Ever had a text or email lost in context? Chances are it happens every day. Not only do we have different languages that we speak, but we have different dialects within those languages. If we humans have a hard time understanding our language, then machines will struggle too. Now with the use of innovative DL models, machines are beginning to understand human language. In the enterprise, these applications are taking shape in three impactful areas: chatbots, text summarization, and voice interfaces.

NLP Chatbots

Chatbots are not new, but the technology has really improved in the last few years. Today you may not even be aware when you’re speaking to a machine (the Turing Test aside). Imagine using all your emails as training data to build an NLP DL model for a chatbot. All those with quick easily repeatable responses could be cleared out of your email box without you having to manually reply. For example, a technical support engineer is commonly asked how to reconnect their email in their email client. The simple answer that covers 85% of the cases has been solved hundreds of times in their email box. The email archive can train a model to solve the problem for use with a chatbot. Not all emails or customer responses should be handled by chatbots, but think of this use to efficiently triage common problems and respond to them quickly.

NLP Text Summarization

Let’s think back to high school. When school lets out, summer reading lists seem like they’ll be easy to tackle over the coming months, but too many fun things can get in the way. So, time flies, summer ends and those books were never opened before the first day of class. Enter CliffsNotes, the invaluable reference guides summarizing classic books for the student procrastinator or those needing a good review. NLP is helping to bring this functionality to the enterprise with text summarization. Now, hours of re-reading notes from a meeting that took place months ago can be reduced to minutes. Or what about career and professional development?

Many people struggle to keep up with the research in their field. Text summarization can help consolidate the high-level points about what’s new and deliver an easily consumable brief. Another use case is reducing the amount of time customer support engineers spend getting up to speed on a critical support issue. Saving minutes or hours for the engineer allows them to more effectively resolve the problem. Thus, NLP text summarization won’t replace reading, but it can help speed up cognition and time to results.

NLP Voice interfaces

Smart speakers and voice assistants are prolific in the consumer space. For example, my 8-year-old uses her smart speaker to help with homework. NLP is at the heart of these emerging voice interfaces and now it’s being deployed in organizations. Remember my doctor visit story a few months ago? Does he really need to carry around a voice recorder only to have his notes dictated and transcribed later? Not with NLP. Once the doctor leaves the patient room their notes can be automatically transcribed, uploaded and made available to their laptop when they need them. Healthcare will benefit greatly from these developments, but voice interfaces will not stop at smart devices for dictation. Enterprise users will continue to request the use of voice as an interface for such tasks as generating sales reports to voice enabled research assistants. I predict there will be a surge of voice interfaces in the enterprise.

Building an Architecture for Natural Language Processing

These three emerging NLP use cases, and many others for AI and DL, require an optimal IT infrastructure to deliver expected user experiences and results. For instance, training NLP models to understand different dialects, voices and tones requires massive amounts of data, perhaps ranging from terabytes to petabytes. And the NLP will also generate even more data. Since AI initiatives start with data first, it’s important to consider the storage required for this most valuable asset.

It’s equally imperative to contemplate your partners for the journey. Dell Technologies and NVIDIA are focused on helping our customers realize the value of their data with innovative AI solutions. Customers trust our expansive portfolios of best of breed hardware and software offerings to deliver high performance and scalable IT Infrastructure from sandbox proofs of concept to large-scale enterprise production. To this end, we’ve delivered Dell Technologies Ready Solutions for AI as well as reference architectures based on Dell EMC Isilon scale-out NAS with NVIDIA DGX-1™ and NVIDIA DGX-2™.  And we’re looking forward to building on these efforts with the release of the new NVIDIA DGX™ A100 system. In the upcoming months, we plan to begin testing, validating and certifying NVIDIA DGX A100 systems with our Emmy Awardwinning Dell EMC Isilon scale-out NAS. Stay tuned for new solutions and reference architectures built around these essential elements of a high performance, scalable AI IT Infrastructure from Dell Technologies and NVIDIA.

Thomas Henson

About the Author: Thomas Henson

Thomas Henson an Unstructured Data Solutions Systems Engineer with a passion for Streaming Analytics, Internet of Things, and Machine Learning at Dell Technologies. He brings experience in Machine Learning Anomaly Detection, Open Source Data Analytics Frameworks, and Simulation Analysis. Thomas is also heavily involved in the Data Analytics community.