Transform 2021

Elevate your enterprise data technology and strategy.

July 12-16

For call center applications, dialogue state tracking (DST) has historically served as a way to determine what a caller wants at a given point in a conversation. But in the real world, the work of a call center agent is much more complex than simply recognizing intents. Agents often have to look up knowledge base articles, review customer histories, and inspect account details all at the same time. Yet none of these aspects is accounted for in popular DST benchmarks. A more realistic environment might use a “dual constraint,” in which an agent needs to accommodate customer requests while considering company policies when taking actions.

In an effort to address this, AI research-driven customer experience company Asapp is releasing Action-Based Conversations Dataset (ABCD), a dataset designed to help develop task-oriented dialogue systems for customer service applications. ABCD contains more than 10,000 human-to-human labeled dialogues with 55 intents requiring sequences of actions constrained by company policies to accomplish tasks.

According to Asapp, ABCD differs from other datasets in that it asks call center agents to adhere to a set of policies. With the dataset, the company proposes two new tasks:

• Action State Tracking (AST), which keeps track of dialogue state when an action has taken place during that turn.
• Cascading Dialogue Success (CDS), a measure of an AI system’s ability to understand actions in context as a whole, which includes the context from other utterances.

AST ostensibly improves upon DST metrics by detecting intents from customer utterances while taking into account agent guidelines. For example, if a customer is entitled to a discount and requests 30% off, but the guidelines stipulate 15%, it would make 30% an apparantly reasonable — but ultimately flawed — choice. To measure a system’s ability to understand these situations, AST adopts overall accuracy as an evaluation metric.

Meanwhile, CDS aims to gauge a system’s skill at understanding actions in context. Whereas AST assumes an action occurs in the current turn, CDS first predicts the type of turn (e.g., utterances, actions, and endings) and then its subsequent details. When the turn is an utterance, the detail is to respond with the best sentence chosen from a list of possible sentences. When the turn is an action, the detail is to choose the appropriate values. And when the turn is an ending, the system should know to end the conversation, according to Asapp.

A CDS score is calculated on every turn, and the system is evaluated based on the percent of remaining steps correctly predicted, averaged across all available turns.

Improving customer experiences

The ubiquity of smartphones and messaging apps — and the constraints of the pandemic — have contributed to increased adoption of conversational technologies. Fifty-six percent of companies told Accenture in a survey that conversational bots and other experiences are driving disruption in their industry. And a Twilio study showed that 9 out of 10 consumers would like the option to use messaging to contact a business.

Even before the pandemic, autonomous agents were on the way to becoming the rule rather than the exception, partly because consumers prefer it that way. According to research published last year by Vonage subsidiary NewVoiceMedia, 25% of people prefer to have their queries handled by a chatbot or other self-service alternative. And Salesforce says roughly 69% of consumers choose chatbots for quick communication with brands.

Unlike other large open-domain dialogue datasets, which are typically built for more general chatbot entertainment purposes, ABCD focuses on increasing the count and diversity of actions and text within the domain of customer service. Call center contributors to the dataset were incentivized through cash bonuses, mimicking the service environments and realistic agent behavior, according to Asapp.

Rather than relying on datasets that expand upon an array of knowledge base lookup actions, ABCD presents a corpus for building more in-depth and task-oriented dialogue systems, Asapp says. The company expects that the dataset and new tasks will create opportunities for researchers to explore better and more reliable models for task-oriented dialogue systems.

“For customer service and call center applications, it is time for both the research community and industry to do better. Models relying on DST as a measure of success have little indication of performance in real-world scenarios, and discerning customer experience leaders should look to other indicators grounded in the conditions that actual call center agents face,” the company wrote in a press release. “We can’t wait to see what the community creates from this dataset. Our contribution to the field with this dataset is another major step to improving machine learning models in customer service.”

Transform 2021

Elevate your enterprise data technology and strategy.

July 12-16

Presented by Western Digital

Automakers continue to roll out vehicles outfitted with advanced driver assistance systems (ADAS) to help cars become safer, smarter, and more enjoyable. Vehicle operations now largely involve electronic systems centered around processors or system on chips (SoCs). Because of this, digital storage has moved into a more important role that is ramping up next-generation vehicles. Simply put, today’s modern cars — electric, self-driving, hybrid, autonomous, you name it — work as they do because of storage.

Data demands are expanding

The automotive market is driving storage growth. According to the NAND Quarterly Market Monitor, Q1 2021, from Yole Développement (Yole), automotive data demand will skyrocket. The market research and strategy consulting company, Yole, estimates the NAND-based automotive storage demand will increase to 4EB in 2021, and should jump to 24EB by 2024, rising further to 78EB by 2027. Autonomous Level 5 vehicles will only be a small percentage of the vehicle shipments in 2027, so the NAND storage growth will be driven mainly by L2-L4 vehicles, which will represent roughly 85% of vehicles shipped. The feature-rich L2-L4 vehicles will provide advanced levels of safety, connectivity, and entertainment — all requiring higher capacity storage. Overall, the anticipation is a 270% CAGR from 2019 to 2027.

Vehicles today are using more data for over-the-air updates, rich mapping systems, high-quality infotainment applications, and advanced driver assist systems that detect roadway activity.

Here’s what automakers should consider when implementing vehicle data storage.

The right storage is important

In the past, vehicles relied on storage for their infotainment systems — mainly for storing the OS and map data. But today, the use cases are very different and the architectures are changing so the type of storage needs to be carefully considered.

Automotive-grade storage is a must. Automotive-grade devices can withstand wider temperature ranges from -40C to 105C and have gone through a more stringent testing flow at the manufacturer making them less susceptible to physical failures.

There are also SD, e.MMC, UFS, and PCIe interfaces available. Which one to choose is in part dictated by the SoC and the interfaces it supports. It also comes down to performance requirements. If high performance is not a requirement, e.MMC will likely be your best choice because you can avoid the challenges of routing and signal integrity that come with high-speed designs.

UFS is the next step up in performance and is gaining popularity as storage capacity points increase with shared storage applications. PCIe is on the horizon and will definitely be an interface used to support the high performance needed in domains and zonal architectures.

Know your workload

Selecting the right capacity and interface are fairly straight forward, but one thing that is often overlooked is how the device will be written to. When vehicle systems are being designed, it is critical to calculate data workload accurately. The endurance of the storage device depends on the amount of data written, measured in terabytes written (TBW), and varies based on the capacity of storage. Every device has a limit, so it is important to determine the data write requirements based on real system workloads.

The write amplification factor (WAF) is an important consideration that is sometimes forgotten when calculating the system workload and TBW.

WAF is the amount of data that is actually written to the NAND versus the amount of data sent to the NAND by the system. There are several factors that affect the WAF but the main one is that data is written in pages and erased in blocks. Simply put, if a lot of random data is sent to the storage device, the pages might not be completely full causing “empty” space that is not fillable unless the data is moved and combined to another page, thus causing additional writes to free up space. Think defragmenting a hard drive to make space and make the HDD perform better. Similarly, this can happen in flash where data may need to be rewritten to different areas to free up storage space. This causes additional writes as data is moved about to maximize storage area and increases the WAF.

Serial data will provide basically a WAF of 1 since it is fills each page completely. What this means is that if 50TBs of data are sent from the host and the WAF is 3.0, the actual TBW to the NAND is 150TB.

There are things that can be done on the host side to reduce this, but if you were to calculate your useful life based on a WAF of 1.0 but it really was 3.0, your actual useful life would only be 1/3 of that and may cause the system to fail unexpectedly. This is just one example of why it is important to understand how your system will be accessing the storage device and analyzing its impact on the life of the product. You may not even be aware of some writes the system is doing. Taking traces and analyzing them will help better understand your work loads.

Digital storage drives innovation

Vehicles today are essentially computers on wheels that will become servers on wheels in the future. With a range of built-in sensors, cameras, and internal networks, they will be able to seamlessly detect everything that’s happening on the road —  including cars, cyclists, pedestrians, and street signs.

As cars continue to be upgraded with more advanced features, they will have smarter functions and technology that will require more storage and generate even more data. This is why proper data storage and knowing your workload is so critical.

Improved vehicle intelligence can help drivers stay steady on the road. ADAS and other features require a lot of application and operating system software that must be stored on reliable storage devices. Additionally, over-the-air (OTA) software updates from manufacturers can be pushed to add new features and benefits, which will increase the need for storage even more. This means that manufacturers will need to plan long term for additional storage space to handle the increased lines of software code.

Next big step

As cars develop more autonomous features and more amounts of data are created, the need for storage capacity and performance to manage all these applications requires strategic planning, careful coordination, and smart execution. We are still in the beginning of the next chapter for the automotive revolution. Within that, data storage’s role in enhancing how people travel will continue to increase.

Russel Ruben is Global Automotive Segment Marketing Director at Western Digital.

Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. Content produced by our editorial team is never influenced by advertisers or sponsors in any way. For more information, contact [email protected].