LOS GATOS, Cali. — As Netflix continues to produce billions of dollars’ worth of original content, it’s easy to forget that the company’s business model is firmly rooted in the delivery of digital content, served with as little friction as possible.

For Los Gatos, California-based Netflix Inc., frequent improvements to how all of its content is delivered — both original and licensed — is not just for subscriber convenience or benefit.

Retaining the company’s 94 million paid subscribers is crucial, but growth is the name of the game and — when reading between the lines of its latest technology improvements — the company has its sights set on emerging markets.

Netflix uploads multiple versions of shows or movies to its cloud servers, encoded in different file sizes. When a subscriber starts watching content, Netflix will know which file to serve, based on the device being used.

A big screen TV on fast home internet service will be served a higher bitrate — the number of bits transmitted per second — more information makes the picture quality better, while someone watching on a cellphone will get a lower bitrate to reduce the amount of bandwidth being used.

Netflix has been trying to refine the way they encode their videos to push significantly better quality video at a lower bitrate, so as more people move to mobile devices, the video they consume won’t take up as much of their bandwidth limits.

But, more importantly, it also means the company can grow its subscriber base in emerging markets where smartphones and data plans are more common than home Internet service. 

“I’m originally from the Philippines, where the main access to the Internet is actually people’s cell phones,” Anne Aaron, Netflix’s director of video algorithms, told a small group of journalists at the company’s headquarters. 

“Every bit counts. So the role of my team is to make sure every bit actually adds to the video quality of what people watch, and our main goal is to have a great viewing experience where you enjoy the TV show or movie at any bit rate.”

Part of the way this is achieved is through efficiency. Netflix’s encoding process was once done on a per-title basis, meaning its algorithms would look at scenes with the most action and use that as a basis for how much to compress the quality of the video.

But Aaron’s team has moved the encoder algorithms to a “per chunk” basis, which would look at one-to-three minute segments at a time, which means they can compress higher quality into smaller bitrate because action moments often aren’t as frequent and the threshold is lower. 

“But why stop there? Let’s go even further and optimize per shot of the video,” Aaron said, adding that Netflix has brought in experts from around the world, including two professors that specialize in encoding, to help make their algorithms even more efficient. 

So now video looks equally as good at half the bitrate — and in some cases, it’s even lower. That drives down the bandwidth costs for subscribers, and potential new users in emerging markets are more likely to be attracted to video that looks good on any device, even at slower speeds.

Chris Ratcliffe/Bloomberg

“Every bit counts”

Language accuracy

Quality video that doesn’t take up a lot of bandwidth is half the battle. Netflix is also innovating when it comes to localization — the subtitles and dubbing done in other languages.

“In 2012, we launched Lilyhammer… in seven languages and 96 language assets,” said Denny Sheehan, Netflix’s director of content localization and quality control. “Cut to (this year) where we’ve launched Iron Fist in 20 languages and we have 572 language assets. And by language assets, I mean subtitles, audio dubs and audio description.” 

For Netflix and Sheehan’s team, the way to nail localization is by focusing on context. In some cases, the company bypasses local companies that offer people for hire and hires translators directly, in case there are questions on things like cultural jokes, voice inflection and other contextual elements that might be missed in a straight translation.

Netflix also uses style guides and glossaries of terminologies or key phrases to make sure there is consistency across shows or movies as well as in the marketing materials and elsewhere in the company. All departments can access an internal Wiki with the up to date style guide.

“To achieve the highest quality we also have to have really high bar for quality control, and so for our originals this is a very through and rigorous approach,” said Sheehan.

“Every subtitle event is gone through by the same quality control evaluator that has done every episode of every season of a series, so the person working on House of Cards season five for Japanese also worked on season one and that way we know that nothing is going to be lost season-to-season.”

To expand into more languages and markets with a high level of accuracy, Netflix launched its own translator program in March called Hermes. Anyone can register and choose a language they speak, then take a quick test.

Those who score in the highest percentiles will be contacted by Netflix and interviewed to become a paid translator. If eventually accepted, they’ll get a unique ID in the system and their history (including accuracy) can be seen both by Netflix or exported to show other companies if someone is looking for a full-time position in the field. 

“Everybody in the process (including quality control) is measured,” said Chris Fetner, Netflix’s director of media engineering partnerships. “If we start to see a trend where we feel like that person is not performing we’ll either coach them up to a new level, up to the level that we expect or we’ll discontinue using them.”

With Netflix’s eyes on new markets to keep its subscriber base growing, these kinds of technological innovations and focus on localization will already be in place during expansion to help bring new countries on board.

“Even if you think about India and places in Latin American, there are places that either the fixed line bandwidth is quite constrained,” said Ken Florance, Netflix’s vice president of content delivery.

“In Africa, India, parts of Asia, parts of Latin America where there wasn’t this huge build out of fixed lines to people’s homes, in a lot of cases some cellular networks are substituting for the last mile. So any of the benefits from a 200 kilobits stream looking great on a cell network in New York City will also be seen and look fantastic on an old copper DSL in Bogota (Colombia).”

Author: Josh McConnell
Source: business.financialpost.com

Categorized in Science & Tech

Cambridge, UK-based True Knowledge released its natural language search engine into private beta yesterday (a story broken by TechCrunchUK).

Like the much-anticipated Powerset, the company aims to give appropriate answers to natural language queries, even if key query terms are not included in the data being indexed. Current search engines are unable to return appropriate results for these queries.

At first glance True Knowledge and Powerset are competitors – but in fact they really aren’t. Powerset is both indexing the web and working to convert natural language queries into database-understandable queries. True Knowledge is only tackling half the problem – the conversion of queries. They are not indexing the web.

Instead, True Knowledge is grabbing data from structured databases, like, for example, the CIA Factbook. In many ways, they are more comparable with Freebase, a startup focused on gathering all the structured data on the web.

The engine certainly looks like it will be useful, though. Results can be returned based on inference of the intended meaning. So a question about if someone is married or not can be answered even if there is no specific structured data about that question. See the demo video to get a better idea of what it is capable of:

http://blip.tv/scripts/flash/showplayer.swf?enablejs=true&feedurl=http%3A%2F%2Ftrueknowledge%2Eblip%2Etv%2Frss&file=
http%3A%2F%2Fblip%2Etv%2Frss%2Fflash%2F473501&showplayerpath=http%3A%2F%2Fblip%2Etv%2Fscripts%2Fflash%2F
showplayer%2Eswf You can also see a number of screen shots with sample queries and results here.

The company has raised “more than” £600,000 from Octopus Ventures. As an interesting aside, founder William Tunstall-Pedoe is good friends with Powerset founder Barney Pell. Both studied at Cambridge together, says Pell, who describes True Knowledge as “cool stuff.”Via Go2Web2.

https://techcrunch.com/2007/11/08/true-knowledge-launches-natural-language-search-engine/

Categorized in Search Engine

airs logo

Association of Internet Research Specialists is the world's leading community for the Internet Research Specialist and provide a Unified Platform that delivers, Education, Training and Certification for Online Research.

Get Exclusive Research Tips in Your Inbox

Receive Great tips via email, enter your email to Subscribe.

Follow Us on Social Media

Book Your Seat for Webinar GET FREE REGISTRATION FOR MEMBERS ONLY      Register Now