SoundHound has attempted to revolutionize the way people interact with mobile devices by delivering technologies with user experiences in sound recognition. As a way to proceed forward, it starts to create a voice-enabled personal assistant to rival Apple's Siri.
It all started in 2015 when SoundHound revealed Hound, a personal assistant that could understand and correctly answer complex questions. The product is based on years of experience and work from its engineers.
SoundHound founder and CEO Keyvan Mohajer explained how he conceived Hound and how he expects to compete against rival digital personal assistants created by tech giants Apple, Google and Microsoft.
Mohajer first came up with the idea for Hound when he was a Ph.D student at Stanford. He believed that the next big evolution in personal computing would come in the form of voice-enabled personal assistants that would not only answer basic questions but would be capable of more or less conversing.
When he first tried to execute the idea, he realized that the technology would require time to develop. And that time is a decade. Pitching the idea to venture capitalist won't be a good idea because most of them would expect a fast return of investment. This was his first challenge.
His first product that came out from that idea was SoundHound, and it was released in a shorter amount or time. SoundHound was Mohajer's product to bridge his bigger ambition.
SoundHound is a music identification app that aims to help users to identify songs from just hearing it. SoundHound is even able to recognize a song from just hearing the user humming it. As the product became popular, Mohajer and his team was given a preview of what would become the software capable of interpreting human speech.
Because people humming a tune aren't doing so in proper key, SoundHound has evolved to fill in many blanks inside a tune to properly identify music syntax. This method was then initiated as a base for Mohajer's bigger goal.
"Most people are not good singers or hummers," said Mohajer. The additional challenge is because "most times when people use the feature, they're drunk."
While SoundHound keeps its pace, the team behind it started to focus on a bigger prize: a voice-recognition software that won't have the drawbacks from the likes of Siri and Google Now. Right from the start, the team knew the software they want to develop, thus eliminating any extra steps that would harm user-experience or time.
"When you speak to Siri, first your voice becomes text then it gets converting to meaning," he says. "It takes two steps, so it's a little bit slower."
Hound made a debut as a voice-recognition software that convert voice to text, to then interprets it simultaneously. This made Hound faster to respond that virtually any virtual assistant in the market. What's more, the team also trained the Hound to be better in understanding complex questions.
This makes Hound to be able to understand more questions, eliminating users the needs for short keyword-based queries.
"This has enabled users to be more demanding and get what they want," he said. "We want people to be power users. No more dumbing down to the level of Siri and Google Now."
In a demo, Mohajer ran through some complex interactions, starting with asking about the day's weather and leading into a variety of logic-based questions. The app answered as it was supposed to, fast and easy.
Mohajer's plan for Houndify starts as an Android version and an iOS version of the app released on March 1st, 2016.
Paving The Roads To Ambition
Hound is marketed as a faster and more accurate personal digital assistant, built with speech-to-meaning recognition technology based on SoundHound's Houndify developer program.
The Hound project first came out in June 2015 when the company began accepting applications for its private beta program. At that time, SoundHound was expecting at least 1,000 testers, but the number grew significantly larger that in just two months - more than 300,000 applications delivered to it from all over the world.
For partners, the company added Expedia, AccuWeather, Xignite, FlightStats, Uber and Yelp.
And as for domains, Hound has more than 100 (a double the number it had when the app was still in private beta) domains to include weather, currency translation, voice dialing and texting, nutrition, music, navigation and directions, reference, video search, unit conversion, word and phrase translation, and sports scores. Mohajer said the number exceeds Siri's and noted that Apple’s assistant launched with 12 domains and took five years to double that.
As Mohajer explained, Hound and other assistant-type apps, "become more useful when you add more domains."
And for how Hound works, it's similar to other virtual assistant available on the market. Users can access it by either pressing the microphone button in the app, or by saying "Okay Hound…". And for those that have privacy concerns, Mohajer explained that, while Hound always listens, nothing is being recorded on the server level.
By making Hound more "human" in recognizing speech, complex questions could be asked. Below is some complex question examples Hound can answer:
- "Show me coffee shops with Wi-Fi" and "Which ones are within walking distance and are open after 9 p.m. on Sundays?"
- "List all Asian restaurants within three miles that aren't Chinese food and that are open between the hours of noon and 8 p.m. on Sundays."
- "What’s the weather?" And "How about tomorrow morning in Austin, Texas?" Or "How about two months from now?"
- "Show me hotels in San Francisco for tomorrow staying for two nights that cost between $200 and $300 per night and are pet-friendly and have a gym and a pool."
- "What's the monthly payment on a $1.2 million home with 20 percent down, paid over 30 years with a 3.95 percent interest rate?"
- And more.
From some of the above examples, we can conclude that the best thing about Hound is that users don't have to speak "geeky". While it's still a work in progress, Hound has more features to be soon implemented in it.
"SoundHound is a deep tech company that isn't forcing the geek world on you. You'd speak like you would to your friend," said Kathleen McMahon, the company's General Manager and Vice President.
Houndify "Everything"
First released as a beta, Hound is impressive as it should be. But it's apparently just a small piece of Mohajer's plan. He wants to "Houndify" everything by making it a platform that allows third-party app developers to integrate it to their apps.
To Houndify is Mohajer's aim to embed not only smartphone apps, but also every devices people have connected to the internet. As Mohajer explained, it's "not just about an app on your phone, it's about anything you interact with."
The competition for voice-enabled digital assistant is just getting fierce. Using natural-language recognition built off of the company’s proprietary platform, SoundHound is taking on Siri, Now, and Cortana, but with initial aims bigger than them.
For Hound's future, Mohajer cited an example where a colleague "houndified" an espresso machine so that while a drink was being made, people could interact with the device and and ask what the weather was like, hear the news, and more. After the company demonstrated the product at the Consumer Electronics Show in Las Vegas, California in 2015, it has some companies willing to use it. Some notable ones was Nvidia for its DRIVE CX platform in car dashboards, and Samsung for its ATRIK platform to power Internet of Things devices.
Houndify platform opened up to allow developers to create products around existing domains, but added that soon they'll also able to create their own domain and monetize it. While Hound is free to use, its business model lies on compensation. So if users request an Uber or book a hotel, the company will receive a percentage of that transaction. And if a user uses its search and buy a song from it, the company will also receive revenue. Products and apps that have been "houndified" will also be monetized in similar ways.
