Amazon Alexa gets smarter and continues to improve

Amazon has been the No. 1 player in the emerging smart home industry. Its voice assistant Alexa and smart speaker Echo line have had the greatest number of users compared with its rivals’. This early lead gives Amazon an edge as nine out of 10 smaller companies will always seek Alexa compatibility when launching smart products.
 
“Amazon’s strategy is to continuously roll out new hardware with Alexa built-in, like Echo, Echo Look, and Echo Show, for different purposes and spaces at home,” said Shawn Lin, Associate Manager of Quanta Computer, who is leading product development for the company’s smart home business venture EQL Technology. “Amazon has been repeating this business cycle in the past year, and has a nice grasp of customers’ behaviors and needs.”
 

EQL’s smart hub Ripple can control Bluetooth devices (sensor, blood pressure monitor, etc), infrared devices (TV, air conditioner, etc) and Wi-Fi devices (IP camera, robotics vacuum, etc). When Ripple is connected with an Echo smart speaker, users will be able to control all these devices with voice. “From a manufacture’s perspective, we need to add the integration,” Lin said.

Alexa is even more popular in the U.S., where the voice assistant is gradually finding wider applications. The recently-built elderly care housing Park Creak Independent Living in Texas is one example. The facility is equipped with complimentary Echo devices so seniors may utter commands to set reminders, call families or the front desk, and get dining and activity information in the community housing.
 

Compared with other new technologies, using voice is more intuitive and so much easier for seniors. The user experience has been very positive, according to Park Creek. Tyllen Bicakcic, co-founder of CT Home, which develops custom Alexa skills for Park Creek, noted, “Alexa had a lot of third party integrations that we were more comfortable with using. Also in the United States a lot more seniors seem to be familiar with Alexa than Google Assistant.”

App interface and business model improvement

The Alexa platform is used by many smart home companies. And improvement on Alexa is constantly made, according to Goodway Technology, which sells a Z-Wave gateway that has been certified by Amazon to work with Alexa.

“After we upgrade our Alexa skill, the companion app will also be upgraded to provide more functions,” said Tony Chen, a project manager at Goodway. For example, the app had been able to just turn on or off a smart plug, but after the skill upgrade, the app can also show how long the plug has been used and how much electricity it has consumed, Chen said.

Another improvement made on Alexa in the perspective of manufacturers is that Alexa skills can now generate revenues, instead of just providing free voice service to customers. Merchants providing Alexa voice service may add new commands to let their customers make purchases.

A company that plans to do so is Zyxel Communications. The company’s mesh Wi-Fi system Multy X is recently integrated with Alexa, allowing users to check Internet speeds and turn off guest Wi-Fi via voice commands for free. In the future, Zyxel may introduce a subscription-based cybersecurity service, which customers may activate via voice commands.

“The free voice service business model is not sustainable, and Google Assistant may follow suit as well,” said Bill Su, Senior Associate VP of Smart Living Strategic Business Unit at Zyxel Communications.

Weakness in natural language processing

Despite Amazon’s leadership position in the smart home space, Alexa has a clear weakness compared with its chief rival Google Assistant. That is its natural language comprehension ability.

According to Goodway Technology, sometimes when a long command is given, Alexa will cut it short and begins to execute command, before the user is finished with his or her sentence. Google Assistant would not make this kind of error, Goodway’s Chen said.

Alexa may also perform rather poorly when the voice command involves specialized domain knowledge. That is the view of Poland-based Infermedica, which created the Symptom Checker skill. Users may open Symptom Checker when they feel unwell. Symptom Checker will ask a series of diagnostic questions to determine the potential cause, and suggest actions to take, for example, seeking medical treatment if necessary.
 

“While Alexa performs well in typical Alexa apps, that is when users speak about shopping, weather, ordering food, sports and so on, it works pretty bad when you speak about symptoms or other health-related topics,” said Adam Radziszewski, an Infermedica data scientist who works on machine learning and natural language processing (NLP).

For example, “dizzy” is usually understood as “Disney.” Roughly half of times when someone tries to report “abdominal pain,” the phrase is scrambled and what you get is “domino game” or “domino pain,” probably because Alexa is often used to order pizza at Domino’s.

Both Cortana and Google Assistant are way better than Alexa in healthcare language interpretation, according to Radziszewski.

Alexa gets more intelligent

Still, Alexa has made some improvement in NLP. When Alexa was first launched, the voice assistant interacted with users mostly through a rather simple straightforward dialogue. A question followed by an answer, and the phrases need to be exact, or Alexa might not understand.

When someone wanted to order tissues, for example, he had to start by telling Alexa that he wanted to purchase something, followed by the type of merchandise, the brand, the amount, etc., all via independent commands.
 

“Today Alexa is able to understand a more complex command, so that a user may order a merchandise using fewer commands,” said Bryan Liu, CEO of Bking Technology, a company that specializes in smart home software integration. A command like “Order a pack of tissue for me tomorrow” consists of three information: the type of merchandise, the amount and order time, and this is a big improvement for Alexa, Liu added.

Also, Alexa can learn the pronunciation and intonation of different users over time, Liu noted. “When someone first purchased an Echo, maybe it can only recognize the trigger word ‘Alexa’ voiced form its owner five out of 10 times. But as time goes on, the number will increase. Also, Alexa’s ability to correctly understand the owner’s commands and intention will increase gradually.”

Hopefully Alexa will become available in more languages and enter more countries. As the Amazon Alexa ecosystem grows and expands into more markets, it will also help other companies and system integrators find niches and opportunities in the smart home industry, Liu concluded.