Meta Platforms Technologies, LLC

Facebook Technologies, LLC

Tali Zvi

Hyunbin Park

Vincent Charles Cheung

Maheen Sohail

Chengxuan Bai

Jiemin Zhang

Andrew Pitcher Thompson

Lior Wolf

Yaniv Taigman

Kent Austin White

David Vazquez

Adam Polyak

Kevin Lin

Xinyue Zhang

The present embodiments relate to automated memory creation and retrieval from moment content items. In some implementations, the automated memory creation and retrieval system can obtain moment content items from user-designated sources with a single user perspective or multiple user perspectives. The moment content items can be assigned tags and arranged in chronological order. The arranged moment content items can be clustered into memory content items based on clustering conditions. Once memory content items are created, they can be arranged into a memory hierarchy made up of connected nodes. In some implementations, the memory content items are also linked together based on similarity in various dimensions in a memory graph. The automated memory creation and retrieval system can receive search criteria for memories from a user interface and provide the user with memories from matched nodes in the memory hierarchy or linked memories in the memory graph.

Automated memory creation and retrieval from moment content items

An automatic memory system can automatically identify and present content items that match a user's natural language (NL) input. The automatic memory system can compute a match score between the NL input and each of multiple potential memory content items. The automatic memory system can compute the match score using a variety of algorithms and/or machine learning models such as an image/NL matching process to get a first match score, a tag matching process to get a second match score, and/or a combination match score from the first and second match scores. The automatic memory system can select one or more of the content items with the highest match score(s). The automatic memory system can provide the selected content items, such as by suggesting them to the NL providing user, automatically displaying or playing them, inserting them into the conversation thread, etc.

Automatic memory content item provisioning

A multimedia query system is described that includes a multimedia capture system configured to capture raw multimedia data comprising at least one of raw video data or raw audio data, a metadata engine configured to extract one or more anchor points of metadata from the raw multimedia data and to store the one or more anchor points of metadata, wherein the anchor points of metadata includes references to respective portions of the raw multimedia data. The multimedia query system further includes a storage engine configured to store the raw multimedia data, a recall engine configured to receive a query and to apply the query to the one or more anchor points of metadata to identify one or more raw multimedia data candidates from the portions of the raw multimedia data, and a query engine configured to generate a response to the query based on the one or more raw multimedia data candidates.

Multimedia query system

Disclosed herein a system, a method and a device for generating a voice model for a user. A device can include an encoder and a decoder to generate a voice model for converting text to an audio output that resembles a voice of the person sending respective text. The encoder can includes a neural network and can receive a plurality of audio samples from a user. The encoder can generate a sequence of values and provide the sequence of values to the decoder. The decoder can establish, using the sequence of values and one or more speaker embeddings of the user, a voice model corresponding to the plurality of audio samples of the user.

Generating a voice model for a user

Growing community of inventors

San Carlos, CA, United States of America

Tali Zvi