Convert enterprise audio data into actionable insight

[TechExperts]

Learn how HPE and NVIDIA’s speech-to-text solution stack optimizes natural language processing.

Enterprises are adopting natural language processing (NLP) technology to extract value and insights from unstructured data. Learn how the joint solution from HPE and NVIDIA lets you deploy NLP at scale.

By Tobore Imarah, Product Manager, HPE, and Ranjini Srinivas, Technical Engagement Manager, HPE

Natural language processing (NLP) and automatic speech recognition (ASR) technologies create a powerful bridge between humans and computers by enabling seamless two-way communications across any platform. Now enterprises can deploy NLP and ASR at scale to convert audio (speech) data into actionable insights.

With the NLP market expected to grow by more than $100 billion by 2028, enterprises are only just starting to unlock the technology’s most powerful advantages.

For instance, ASR (also known as speech to text, or STT) technology is already being used by emergency services to monitor calls and transcribe audio conversations. But the potential is far greater. For example, police departments can run sentiment analysis to triage the most pressing emergencies and determine the appropriate professionals to handle different crises.

Despite this potential, enterprises are still looking for effective ways to make their data available for natural language processing. In fact, up to 90% of the world’s data is unstructured and locked in scanned documents, email, video, audio, and social media.

Businesses are sitting on spoiling data with limited ways to extract value from it. They need a solution stack that can optimize the collection, processing, and analysis of audio data to capitalize on the ASR opportunity. Such an endeavor would involve practical challenges, like deploying ASR-based applications at scale while ensuring high accuracy, addressing data security and privacy concerns, enabling interactions in many languages, translating industry-specific jargon, and enabling real-time responses. 

That’s where the collaboration between HPE and NVIDIA comes in. HPE’s AI platform powered by NVIDIA enables businesses to accelerate the deployment of NLP applications through a fully validated reference architecture. This joint NLP and ASR solution, enables enterprises to deploy powerful ASR technologies with the highest possible accuracy at the lowest possible cost.

Next-gen automatic speech recognition

At GTC 2022 this week, you have the opportunity to learn about the latest capabilities of  NVIDIA Riva, the GPU-accelerated, fully customizable software development kit for building and deploying speech AI – ASR and Text-to-Speech (TTS) — applications. Riva allows every stage of the speech AI pipeline to be customized to solve customers’ unique challenges. Its state-of-the-art out-of-the-box models – with optional fine-tuning for dialects, accents, noisy environments, and industry-specific jargon – enables highly accurate ASR when run on a Kubernetes-based NLP reference architecture. 

Why Kubernetes for NLP workloads?

Containerization ensures consistent deployment across datacenters or cloud environments, minimal disruption to users, and automatic scaling depending on demand. Moreover, Kubernetes abstracts the infrastructure layer, allowing NLP workloads to take advantage of containerized GPUs and standardize data source ingestion. When deploying containerization at this scale, enterprises need a way to manage their Kubernetes environment to further optimize NLP workloads.

HPE is proud to team with NVIDIA to facilitate easy adoption of NLP with Riva ASR and help enterprises build conversational applications in the cloud, on-prem, and at the edge.

Here’s how:

First, HPE Ezmeral Runtime Enterprise provides enterprise-grade container management for Kubernetes-based language processing. Designed to run both cloud-native and non-cloud-native applications at scale with persistent data, Ezmeral helps data scientists support multiple enterprise-grade containerized AI/ML applications via multi-tenancy. This enables organizations to deploy NLP and ASR workloads faster and greatly accelerate time to value.

HPE Ezmeral also features multiple levels of built-in security controls to integrate with identity providers such as AD/LDAP, single sign-on, and SAML integration. Additionally, clusters and applications can be further secured by using strongly attested identities provided by SPIRE for authentication. And users can enjoy intelligent traffic shaping, load balancing, canary rollouts, and A/B testing of NLP application microservices through HPE Ezmeral’s built-in service mesh. 

Next, HPE ProLiant servers enable the high-performance computing applications required to turn audio into insights. HPE systems that are NVIDIA-Certified bring together HPE servers and NVIDIA GPUs in optimized configurations that are validated for performance, manageability, security, and scalability, and backed by enterprise-grade support.  IT professionals can even get help architecting, training, and project managing ASR solutions with HPE PointNext services.

Finally, through the open and secure HPE GreenLake edge-to-cloud platform, you can get a unified cloud experience and operating model for your AI/ML workloads across data centers, co-locations, and the edges. HPE GreenLake is the premier provider of the cloud outside the public cloud and is fully invested in making NVIDIA AI available as a service. HPE recently announced that the NVIDIA AI Enterprise software suite is now available on HPE GreenLake, reducing risk and optimizing costs by delivering NVIDIA AI as a service.  By simplifying IT operation, HPE GreenLake helps organizations manage performance, cost, security, and compliance of data across the hybrid cloud.  

Optimized solution stack

With the combined power of NVIDIA AI and HPE, enterprises can build highly scalable and accurate ASR and NLP solutions. Here are just a few examples: Emergency services departments can easily triage calls and send help faster and potentially save lives. Contact centers can understand customer sentiment and improve first-call resolutions for a better customer experience. And financial services organizations can detect fraud and tailor customer experiences.

This exciting joint solution of NVIDIA Riva ASR with robust HPE NLP software and Kubernetes-based NLP-enabling hardware will allow organizations of every stripe to unlock new potential in untapped data streams.

Learn more about HPE and NVIDIA’s joint solution for ASR technology.

Check out other HPE panels at GTC:

Accelerate AI by Teaming up with the Right Crew and the Best Infrastructure, with Evan Sparks [A41399]

Accelerating AI Edge to Cloud with NVIDIA and HPE GreenLake with Jeroen Bronkhorst and Terry Chiang [A41396]

---

Meet our Tech Experts bloggers

Tobore Imarah, Product Manager, HPE

Tobore is a product manager in the Mainstream Compute Workload-Solutions team focusing on artificial intelligence. His charter is to bring solutions to market that are optimized for performance and efficiency. Tobore received his B.S. in Industrial and Systems Engineering from Kennesaw State University and an MBA from Western Governors University.

Ranjini Srinivas, Technical Engagement Manager, HPE

Ranjini is a technical engagement manager for the HPE Ezmeral ISV ecosystem. She works closely with ISVs to facilitate their product validation on HPE Ezmeral Runtime. Ranjini received her MS degree in computer science from the University of Nebraska-Lincoln.

Insights Experts
Hewlett Packard Enterprise

twitter.com/HPE_AI
linkedin.com/showcase/hpe-ai/
hpe.com/us/en/solutions/artificial-intelligence.html