Column At the annual spectacular of crap that we optimistically term the Consumer Electronics Show, I found myself locked into a room with Alexa. And yes, disembodied AI servant though she may be, she still filled every corner of a rather spacious room with all of the various products now connected to and enabled by Alexa - everything from alarm clocks to refrigerators to automobiles to - kitchen taps? Kohler Konnect (klever) ties a connected kitchen water tap into Alexa, because... "Alexa, pour me 500ml of water." The Kohler Konnect is a kitchen robot of sorts, very good at doing exactly one thing - pouring out precisely the amount of water you need. Is that enough to justify Alexifying a kitchen?

Back in September, Amazon held a massive Alexa event, and one of the many, many products it announced was a clock. The idea was that you'd use a connected Alexa device to control Echo Wall Clock with your voice, and set timers and reminders -- it doesn't have its own speaker or microphone. However, only a month after starting to ship the clock, Amazon has temporarily stopped selling the product because of connectivity problems. "We're aware that a small number of customers have had issues with connectivity," Amazon told The Verge. The company is working on a fix, and plans to start selling the item again "in the coming weeks."

The ultimate streaming box for gamers and audiophiles is back on sale for a good price. You can pick up Nvidia's Shield TV with remote from Amazon for $149 right now. That's $10 higher than the sale price we saw in early November, but it's still lower than the usual $170 to $180 for this box, which doesn't go on sale very often. The Nvidia Shield TV is an outstanding living room device. At its core, it's an Android TV box, but the Shield TV can do so much more than simply stream Netflix and Hulu.

Voice technology is changing how we surf the net, buy products and interact with our devices. Already, we can search for coffee shops while we're driving and order pizza from our home automation devices. In the coming years, we'll be able to talk to our refrigerators, ask our fitness trackers questions, have conversations with our TVs and so much more. As people ditch their touchpads and keyboards, businesses will have to adapt. Smart leaders and entrepreneurs have already started using voice technology to their advantage.

If there's one common theme at this year's CES 2019 – the Las Vegas tech expo we've been covering all week – it's talking. Not just chatting on the show floor (though there's been plenty of that too). By talking we mean the sheer avalanche of devices arriving with support for Amazon Alexa, Google Assistant and so many other competing voice assistants. In fact, everything from motorcycle helmets, water bottles and phone chargers to doorbells, pianos and toilets at least now has the option of being controlled by voice. "Getting rid of the remote control is one of the strongest use cases for voice technologies," says Mark Lippett, CEO at XMOS, whose farfield voice tech is used in soundbars, Freebox and Skyworth TVs to allow them connect to Alexa.

But: "Vulnerabilities in voice assistants like Microsoft's Cortana and Amazon's Alexa are perfect examples of how the rush to simplify can cause complicated problems down the road".

We propose a novel end-to-end Aspect-based Rating Prediction model (AspeRa) that estimates user rating based on review texts for the items and at the same time discovers coherent aspects of reviews that can be used to explain predictions or profile users. The AspeRa model uses max-margin losses for joint item and user embedding learning and a dual-headed architecture; it significantly outperforms recently proposed state-of-the-art models such as DeepCoNN, HFT, NARRE, and TransRev on two real world data sets of user reviews. With qualitative examination of the aspects and quantitative evaluation of rating prediction models based on these aspects, we show how aspect embeddings can be used in a recommender system.

Collaborative filtering often suffers from sparsity and cold start problems in real recommendation scenarios, therefore, researchers and engineers usually use side information to address the issues and improve the performance of recommender systems. In this paper, we consider knowledge graphs as the source of side information. We propose MKR, a Multi-task feature learning approach for Knowledge graph enhanced Recommendation. MKR is a deep end-to-end framework that utilizes knowledge graph embedding task to assist recommendation task. The two tasks are associated by cross&compress units, which automatically share latent features and learn high-order interactions between items in recommender systems and entities in the knowledge graph. We prove that cross&compress units have sufficient capability of polynomial approximation, and show that MKR is a generalized framework over several representative methods of recommender systems and multi-task learning. Through extensive experiments on real-world datasets, we demonstrate that MKR achieves substantial gains in movie, book, music, and news recommendation, over state-of-the-art baselines. MKR is also shown to be able to maintain a decent performance even if user-item interactions are sparse.

In this paper we consider an online recommendation setting, where a platform recommends a sequence of items to its users at every time period. The users respond by selecting one of the items recommended or abandon the platform due to fatigue from seeing less useful items. Assuming a parametric stochastic model of user behavior, which captures positional effects of these items as well as the abandoning behavior of users, the platform's goal is to recommend sequences of items that are competitive to the single best sequence of items in hindsight, without knowing the true user model a priori. Naively applying a stochastic bandit algorithm in this setting leads to an exponential dependence on the number of items. We propose a new Thompson sampling based algorithm with expected regret that is polynomial in the number of items in this combinatorial setting, and performs extremely well in practice. We also show a contextual version of our solution.

For example, Netflix engages its audience by making customized content recommendations to each user. The company reported in 2012 that three-quarters of viewings originated from such suggestions. To make personalized recommendations at such scale, the company leverages an autonomous, integrated learning system that receives feedback (in the form of user ratings or viewer behavior) and updates suggestions accordingly. And it deliberately introduces variation into its recommendations, allowing new behaviors to emerge and letting the data speak for itself. For example, Amazon's dozens of data science systems are closely integrated, which allows decision engines to react to new data immediately and consistently.