DeepText: Facebook's attempt to revolutionise user interaction

Facebook as a platform has integrated numerous features to make its platform better for users. One of the major areas that the company has focused on is making the platform smarter through the addition of intelligent AI bots, helping users to interact effortlessly for all their requisites.   

Now, the social media giant has announced that it is working on a new machine learning technique, dubbed DeepText, which is a AI engine working on the principle of Neuro-linguistic programming (NLP).

According to a blog post on the company’s developer forum, a company representative said that deep learning-based text understanding engine has the capability to understand the textual content of several thousands posts per second, spanning more than 20 languages. And that too, with ‘near-human accuracy’.

Contrary to the numerous updates that deal with minor site improvements, this feature will drastically change how users interact with AI engines. It will not only enhance user experience but also make Facebook a powerful search engine.

Using this intuitive machine learning technique, users will be given numerous options based on their posts, irrespective of how it is framed. DeepText has the capability to understand the core subject of a post; the text understanding capability is varied, ranging from sports, names of celebrities, match stats, and all information relevant to a post.

However, the main aim of the project is to tech the AI engine to understand, differentiate, and learn new words and slangs. The blog post also pointed out that regular NLP techniques are not sufficient for solving tricky language challenges and deep learning is imperative to understand text better across myriad languages.

“There are also variations within each language, as people use slang and different spellings to communicate the same idea. Using deep learning, we can reduce the reliance on language-dependent knowledge, as the system can learn from text with no or little pre-processing. This helps us span multiple languages quickly, with minimal engineering effort,” the blog said.

The company is also beginning to use high-accuracy, multi-language DeepText models to help people connect with the right tools for a task. For instance, if the AI engine detects the subject of a post, it will show the user relevant details and options to the user for getting the work done faster.

As for the future roadmap of DeepText, Facebook will continue to enhance the feature and applications tied to it, in collaboration with the company’s research team. Some of the goals are better understanding of people’s interests and sentiments, joint understanding of visual and textual content, and new deep neural architecture for better word learning.

If this project sees the light of the day, it will help the social media giant mould unstructured data into valuable information that can not only help understand its users better but also to improve its own business.