Facebook knows a heap about you through the collected data it has on everything you Like, share and engage with on the platform, every day. But the value of that data is only realized when Facebook’s able to put it into context – by using those insights to fuel the News Feed algorithm for example, Facebook can deliver a much more customized and personalized experience, because it knows what you’re likely to be most interested in.
On this front, Facebook’s always looking to better utilize its data sources to customize and streamline your on-platform experience. And their latest efforts on this front will do just that – though they might also make privacy advocates a little uneasy.
In a new post on Facebook’s Code blog, the engineering team have outlined what they’re calling ‘DeepText’. DeepText is a new system that enables Facebook to analyze the billions of conversations that are shared across The Social Network every day to provide better contextual matching to associated functions on the platform.
For example, if you were to write ‘I need a ride’ in Messenger, Facebook, through DeepText, will now be able to analyze what you’ve written and suggest that you book a Lyft or Uber with a built-in prompt, direct from the message thread.
Or if you were looking to sell something on the platform, DeepText will be able to analyze the content of your post and suggest that you use the other, available selling options on Facebook to improve your chances of making the sale.
The idea behind the process is that it will improve people’s on-platform experiences by increasing awareness of the relevant, available options, streamlining your intended actions, while also raising awareness of Facebook’s tools.
In order for this to be effective, of course, the DeepText system needs to utilize some advanced contextual parameters to better understand the nature of each request – it needs to be able to tell the difference between ‘I just got a ride’ to ‘I need a taxi’, for example, and it needs to be able to differentiate based on regional colloquialisms and slang.
To assist with its ongoing development, Facebook’s grouping common language sources use via “semi-supervised labels using public Facebook pages”.
“It's reasonable to assume that the posts on these pages will represent a dedicated topic — for example, posts on the Steelers page will contain text about the Steelers football team. Using this content, we train a general interest classifier we call PageSpace, which uses DeepText as its underlying technology. In turn, this could further improve the text understanding system across other Facebook experiences.”
As such, Facebook already has a huge data set of contextual parameters, helping it to advance the system’s learning capacity at a much faster rate.
In many ways, the system’s ingenious, but, of course, it does mean that Facebook will need to be tracking your every comment in Messenger, your every post on the platform. Which they were already doing anyway, but this process just makes it more overt, which will no doubt raise concerns amongst some user groups.
And while these initial use-cases for Deep Text are promising, what could be more interesting is how the same data could be used to fuel additional functionalities. As noted by TechCrunch, DeepText could, eventually, be used to help detect and stamp out harassment and abuse. DeepText could also add a significant new element to Facebook’s News Feed algorithm – if DeepText is able to generate a higher level of understanding about what people are talking about and interested in by analyzing the context – not just the content – of their every post, that’ll enable Zuck and Co to understand even more about what you, and others like you, want to see more of.
Such advancements could also have significant implications for ad targeting, especially as the accuracy of their contextual matching improves.
And that, of course, is before you even consider pairing the technology with Facebook’s other advancements in the areas of image recognition and the AI processes which it’s using to fuel its new Messenger bot service. As DeepText evolves, so too, theoretically, will the capacity of those new Messenger bots to better understand every question and query thrown at them, improving their functionality and utility.
But even if it’s not expanded more widely, the option of being able to immediately connect to relevant Facebook functions will no doubt get more people using them – if you can get the service you want in one click, without any extra effort on your part, why wouldn’t you click through?
As more data is created, the need for tools and systems which are able to translate it into actionable insights and functions also increases. Through this latest development, Facebook underlines its position as a leader in this field, and the one most likely to benefit, both in the near and long term.