meChat: In-Device Personal Assistant for Conversational Photo Sharing

Abstract

It is still challenging to search in-device photos without the sacrifice of privacy or poor latency, while photos have been an essential medium of the communication in messaging applications. In this paper, we propose a novel in-device personal assistant for conversational photo sharing, called meChat. By understanding the user's conversations and in-device photos, meChat intelligently searches in-device photos that are semantically relevant to the conversation context. Notably, meChat works in a stand-alone, privacy-protecting manner without sending out any in-device personal content. The scenario-based user studies show that meChat efficiently searches highly relevant in-device photos with low perceived latency and energy consumption.