Google’s ML Kit SDK adds Smart Reply and Language Identification APIs
Google’s Machine Learning (ML) Kit is an SDK that incorporates years of Google’s work on machine learning into a Firebase package. The goal with this ML Kit is to allow mobile app developers on both iOS and Android to benefit from the real-world uses of ML without needing to understand how the algorithm works. Google’s Firebase SDK for Machine Learning began with five APIs, namely, for text recognition, face detection, barcode scanning, image labeling and landmark recognition. And now, Google has announced that the API is expanding into Natural Language Processing (NLP) to include Language Identification and Smart Reply.
Natural Language Processing is a category of ML that deals with analyzing and generating text and speech. These two new APIs fall within the scope of NLP, they both work fully on-device and are available on the latest version of the ML Kit SDK.
A lot of apps have functionality that revolves around the user specifying the language they are using, which in turn affects features like spell check, translation, and Smart Reply. You can ask users to specify the language, or you can also use the new Language Identification API which can recognize text in 103 different languages. The API requires only a few words to make an accurate determination, and can typically provide a response within 1 to 2 ms.
Smart Reply API aims to provide contextual text response suggestions based on the previous messages in the conversation. This API provides suggestions based on the last 10 messages in a conversation, but it can also work even with one past message. The API is completely stateless and runs fully on-device, so no message history is retained anywhere or sent to a server. The Smart Reply API is launching with English language support first.
Google has worked closely with its partners at textPlus to ensure that the Smart Reply API is ready for developers to incorporate the same. You can see the app utilizing the API to provide contextual response suggestions.
App developers can utilize both these new APIs into their apps. Please refer to the documentation for more information.