Google today introduced new natural language processing services for ML Kit, its solution for iOS and Android developers to bring AI to their apps with Firebase.
Smart Reply will suggest text responses based on the last 10 exchanged messages. The model runs entirely on device and does not store text data.
Also introduced today, Language Identification can identify a language being spoken in a string of texts, and can discern 103 different languages.
ML Kit made its debut in May 2018 at the I/O developer conference, but at launch it was focused primarily on computer vision services such as object character recognition (OCR), face detection, barcode scanning, and image labeling, which can run on-device or in the cloud.
A beta version of face contours to recognize over 100 detailed points of a person’s face was added for the Face Detection API last fall.
Like other ML Kit services, Smart Reply and Language Identification will be made available for free to developers.
ML Kit deploys premade AI models using TensorFlow Lite. TensorFlow Lite 1.0 and TensorFlow 2.0 made their debut last month. ML Kit for Firebase can also be used to deploy custom TensorFlow Lite models in Firebase.
Additional ML Kit features will likely be shared next month at the I/O developer conference, Google said in a blog post today.