User context

Meta’s January 2022 patent proposes working out user context partly from communication intent—what a person is trying to say1. The Meta patent also lists profile information, device location, calendar information and information about external events as possible inputs1. Its wording allows these inputs to be used in different combinations1.

Meta’s graphical-element patents illustrate user context with a person who intends to attend a birthday party at a particular address, with two friends also planning to attend345. These Meta patents also describe current interests, such as favorite artists, movies, music, books, sports, places and foods345. In the Meta patents, those interests may be inferred from profile details, searches, biographical or demographic information, events and group memberships, rather than only from an explicitly stated preference35.

Meta’s graphical-element patents treat user context as one consideration when recommending sticker-like graphics for messages, not as the sole selection rule145. The Meta patents also allow promotion information to influence recommendations, giving promoted graphics greater visibility145. Meta’s patents further describe taking account of whether the person already owns a graphic, distinguishing contextual relevance from ownership-related selection45.

Meta’s May 2021 patent proposes another way to make suggestions relevant to a person’s immediate activity: responding to an action such as reading a post or receiving a message2. The Meta patent combines the content associated with that action, related content and the person’s profile into a comparable representation called a “feature vector”2. The Meta patent then compares that representation with representations of possible recommended content, selects suggestions and sends instructions to display them in response to the original action2.