Facebook unveils a new "text understanding engine" that will be able to comprehend content with "near human accuracy" and to learn 20 languages.
According to CNet, languages could be crucial in Facebook's strategy to cover a massive global market. In this context, the new DeepText system could help open doors to a much smarter global AI.
Bot-based understanding and artificial intelligence could take another leap with DeepText. According to Facebook's detailed post published on Wednesday, June 1, DeepText is a deep-learning artificial intelligence system that is able to understand the content of several thousand posts per second.
Facebook explained that the aim of DeepText is to figure out in a more intelligently and contextually way what is being said in any post at any time. The system will provide a global better understanding of what's being said.
Facebook also said in its post that it is important for DeepText to understand as many languages as possible because the community on Facebook is truly global. The social media network has already a base of 1.65 billion users and 84 percent of them are outside North America. In this context, it is clear that language comprehension requires AI systems such as DeepText to exist.
PCMag reports that Facebook's new DeepText AI may help users to sell used furniture, call an Uber or ensure they never miss their favorite celebrity's posts. The project aims to help improve users' experiences on the social media network.
The DeepText AI project is led by Facebook engineers Aparna Lakshmiratan, Ahmad Abdulkader and Joy Zhang. Based on the DeepText AI system's analysis, Facebook will be able to decide whether to filter out undesirable content like spam or to surface more of the content that people want to see.
The DeepText AI is based on deep learning, using "word embeddings" to preserve the semantic relationship among lexicon. This system can also apply across multiple languages.
