Today in the US, Facebook released Deep Text, an AI engine built to understand the meaning and sentiment behind all users posting messages and announcements on Facebook. In a statement it said, Facebook said it is working on a build system that can recommend content that people might be interested in and get rid of spam.
This sounds like a small improvement, but it actually has the potential to transform the social network that most of us use every day into what we use every day: a powerful search engine.
"We want to use Deep Text to categorize content on Facebook to make it easy to find the right content to recommend to users," said Hussein Mehanna, director of engineering for Facebook's machine learning team.
Facebook's search field is indeed not like the entire global network that Google is targeting, but it is still huge. More than one billion people use Facebook every day, and the network has trillions of status updates, event invitations, photo albums and videos (on their servers). Facebook is embracing a growing mountain of information that connects people with similar interests more effectively, sells more ads, and helps people find what they are looking for.
Previously, Facebook had used demographic information shared by users (either directly or through its brand interaction on the site), but now most text-based information on Facebook servers is unstructured, meaning that Facebook does not Know what the user intended to post a message. Deep Text can help classify messages and provide meaning for all text, turning all unstructured data into information that users can use and search. “If we can understand the words, we can connect people together and share them in many different ways,†Mehanna said.
With this new project, Facebook is basically building the ability to track all the information in its network, just as Google can crawl information and index across the web. This means that users can find useful information more easily in those trillions of messages, just as Google has begun to use artificial intelligence to understand our problems and sort the search results correctly.
Last year, Facebook upgraded its search function to include more search results, such as your search for "barbecue", the results will include photos of barbecues posted by your friends, showing your nearby rotisserie based on your location, involving barbecue International major news stories.
But in theory, when Deep Text figure out what's behind your friend and store posting barbecue news, it can take the search a step further and find the most useful results for you. If you want to know where to eat super delicious barbecue, Deep Text can analyze your friends' news about the barbecue and use it to provide you with restaurant recommendations. If you are looking for information about the health of the barbecue, it can provide the latest relevant news.
Also based on neural networks, but Deep Text is not like other systems designed to understand written languages. Facebook says it can understand the meaning of thousands of messages per second, "close to human accuracy," including 20 languages. The system attempts to understand the semantic relationship and similarity between words, which means that it recognizes that "brother" and "bro" are usually used in similar situations. Deep Text has been trained in a lot of data, meaning it can also understand cross-language similarities, so it doesn't make any difference between "Happy Birthday" and "happy birthday".
The company said that Deep Text is already driving some aspects of Facebook. For example, some chat bots on Facebook Messenger know that they need a taxi based on what people say. For example, if someone sends "I need a ride" to someone else, the robot will interject and ask if they should call a taxi for them. However, if the message is "I just got on a taxi," the robot knows that humans don't need to call a taxi. Mehanna does not want to confirm whether Facebook uses Deep Text on its AI-based virtual assistant M, but it is said that Deep Text is "progressively widely used on Messenger."
At the same time, Deep Text will also appear in other parts of Facebook. For example, if someone posts a message saying they want to sell something, Facebook's business team will automatically get all the information, items, prices, locations, and allow the team to advertise to help people sell items more easily.
Facebook said it plans to build more training data for Deep Text on millions of pages that users have created. For example, the team will use the Pittsburgh Steelers “page to learn more about how people talk about American football and the Steelers team. All of this data will help the team build an AI system that understands the way humans talk online, the connection between words and sentences. .
With this information, Facebook will be able to tag and categorize all the messages posted on its website, making it easier to find things. If Facebook can tell you that your friend, or your friend's friend, is selling what you are looking for, then you don't need to leave Facebook to search for things on Craigslist (large free classifieds site). If Facebook can tell you that the people you know are going to drive to the concert, you don't have to take a taxi.
The recent strong earnings results and CEO Mark Zuckerberg hope to connect the entire world through connected drones and cure every disease, and many people have been saying that Facebook is the new Google.
Facebook's powerful ability to better analyze the data that is delivered to its data center every day can only help keep Facebook users' search needs, rather than relying entirely on Google. But it's worth considering: Facebook provides us with personalized search results, albeit useful, which puts us in a more closed network, formed by our geography, population, affinity and beliefs. Although, to a certain extent, Google does, it does, at least, it searches the entire network first.
Trailer Type Diesel Generator,Trailer Diesel Generator,Trailer Diesel Genset,Mobile Diesel Geneator
Jiangsu Vantek Power Machinery Co., Ltd , https://www.vantekpower.com