Skip to main content

Facebook’s translations are now powered completely by AI

Every day, Facebook performs some 4.5 billion automatic translations — and as of yesterday, they’re all processed using neural networks. Previously, the social networking site used simpler phrase-based machine translation models, but it’s now switched to the more advanced method. “Creating seamless, highly accurate translation experiences for the 2 billion people who use Facebook is difficult,” explained the company in a blog post. “We need to account for context, slang, typos, abbreviations, and intent simultaneously.”

The big difference between the old system and the new one is the attention span. While the phrase-based system translated sentences word by word, or by looking at short phrases, the neural networks consider whole sentences at a time. They do this using a particular sort of machine learning component known as an LSTM or long short-term memory network.

The benefits are pretty clear. Compare these two examples from Facebook of a Turkish-to-English translation. The top one comes from the old phrase-based system, and the bottom one from the new system. You can see how taking into account the full context of the sentence produces a more accurate result.

“With the new system, we saw an average relative increase of 11 percent in BLEU — a widely used metric for judging the accuracy of machine translation — across all languages compared with the phrase-based systems,” the company said.

When a word in a sentence doesn’t have a direct corresponding translation in a target language, the neural system will generate a placeholder for the unknown word. A translation of that word is searched for in a sort of in-house dictionary built from Facebook’s training data, and the unknown word is replaced. That allows abbreviations like “tmrw” to be translated into their intended meaning — “tomorrow.”

“Neural networks open up many future development paths related to adding further context, such as a photo accompanying the text of a post, to create better translations,” the company said. “We are also starting to explore multilingual models that can translate many different language directions.”


Labels:

Comments

Post a Comment

Popular posts from this blog

So this is basically / Asi que esto es basicamente... [SPANISH TEXT]

Si amigos, basicamente la idea del blog fue introducir a todos en el mundo de la tecnologia y hacer que esta no fuera tan "compleja" o "complicada" para todos. Ultimamente no hago reviews propios, ya que me tomo la molestia de elegir buenas noticias (que considero) para su placer informativo (bueno, las visitas me dicen que lo estoy haciendo bien) Pero, y si algun dia llegase a terminar todo? Regalar el dominio? Vender el blog? Nah, muchas veces me lo he preguntado pero... por algo senti el deseo de escribirles, desde mi misma mano y tecla, porque esto es lo que me apasiona: la tecnologia, la programacion, el llevar todo niveles superiores, exponenciar mi capacidad de analisis. De esto se trata todo, esto es basicamente el alma del blog: tecnologia. Actualmente me encuentro en otra ciudad, desde hace ya 1 mes. Las cosas han estado normales, pues dentro de lo que alguien podria definir de "normal". Gracias a Dios no me hace falta lo basico, desafortunad...
Post a Comment
Read more

Child-friendly Galaxy Tab 3 Kids listed in Korean brochure

We're no experts in Korean back-to-school literature, but it looks as if one retailer has tipped Samsung's plans a little early. If the documents above are legitimate, then the company will launch a kiddie-focused Galaxy Tab in short order. The Galaxy Tab 3 Kids is said to be an 8.5-inch slate with a 1.2GHz dual-core CPU, a 1,024 x 600 WSVGA display, 8GB storage, 1GB RAM and Jelly Bean. The company has also seen fit to include 802.11 a/b/g/n WiFi, Bluetooth 3.0, a microSD card slot (no word on capacity) and a 4,000mAh battery. One thing that lends weight to the listing is that the device's model number is SM-T2105, which evleaks tersely described as a "Galaxy Tab for children" a month ago. There's more pictures over at the source, but not a single spec saying that this new device is resistant to jam-smeared fingers. Source: ENGADGET
Post a Comment
Read more

The Ford Fiesta 2011 Was the Budget Hacker’s Dream (And No One Noticed)

The Ford Fiesta 2011 Was the Budget Hacker’s Dream (And No One Noticed) If you ever drove a Ford Fiesta 2011 SE and felt like it had hidden potential, you weren’t wrong — it was a software-defined vehicle before that was even a buzzword . While most saw it as a humble economy car, tinkerers and enthusiasts quickly discovered that the Fiesta was actually modular, reprogrammable, and surprisingly future-proof . With the right tools (and a bit of nerve), you could unlock features typically reserved for higher trims, all with minor hardware tweaks and some clever software work. Here’s a deep dive into the hidden arsenal of the 2011 Fiesta — and why it deserves a cult status among modders. The Secret Weapon: Shared Architecture Ford built the Fiesta using a highly modular electronic architecture . Many trims — from the base SE to the Titanium — shared the same PCM, wiring harnesses, and core modules . That meant you could: Add hardware from higher trims (like steering wheel...
Post a Comment
Read more