SECURELY INTEGRATING STORED ACCOUNT DATA WITH EXTERNAL WORKFLOWS AND LARGE LANGUAGE MODELS

    公开(公告)号:US20250156681A1

    公开(公告)日:2025-05-15

    申请号:US18507630

    申请日:2023-11-13

    Applicant: Dropbox, Inc.

    Abstract: The present disclosure relates to systems, non-transitory computer-readable media, and methods for using a large language model and digital content items (e.g., digital files) stored in a content management system to generate workflow outputs for external computing platforms. For instance, in some embodiments, the disclosed systems receive a workflow request that includes a natural language description of an objective to be accomplished via a workflow. The disclosed systems determine an action plan for completing the workflow and executes the action plan by accessing digital content items (e.g., digital files) stored in a content management system and extracting or generating data from the contents of the digital content items. Further, the disclosed systems use a large language model to generate a workflow output that is provided to an external computing platform for use.

    SEMI-AUTONOMOUS, AI-POWERED HIGH-ORDER OBJECTIVE DISSECTOR AND TRANSPARENT PROGRESSION SYSTEM

    公开(公告)号:US20250156154A1

    公开(公告)日:2025-05-15

    申请号:US18507646

    申请日:2023-11-13

    Applicant: Dropbox, Inc.

    Inventor: Patrick Perey

    Abstract: The present disclosure relates to systems, non-transitory computer-readable media, and methods for decomposing high-order objectives to implement automated or semi-automated changes to software applications. In particular, in one or more embodiments, the disclosed systems receive, from a client device, a high-order objective to be achieved within a computing system. In addition, in some embodiments, the disclosed systems determine, from the high-order objective, a set of sub-processes available to the computing system that combine to accomplish the high-order objective. Moreover, in some implementations, the disclosed systems generate, for a sub-process from among the set of sub-processes, a logic breakdown comprising a description of the sub-process and its predicted effect toward the high-order objective. Furthermore, in some embodiments, the disclosed systems provide the logic breakdown for display via the client device.

    ACCESSING CONTENT ITEMS FOR MEETINGS THROUGH A DESKTOP TRAY

    公开(公告)号:US20250150496A1

    公开(公告)日:2025-05-08

    申请号:US19016823

    申请日:2025-01-10

    Applicant: Dropbox, Inc.

    Abstract: Systems and methods for creating and accessing content items are provided. In some examples, a method can include receiving a user selection of an interface element located in a persistent user interface of an operating system desktop of a client device associated with a user account, identifying, based on a query to a meeting service, a meeting associated with the user account, displaying, in response to the user selection of the interface element, a set of user options for the meeting, the set of user options comprising an option to access a content item for the meeting, receiving a user selection of the option to access the content item for the meeting, and accessing, in response to the user selection, the content item.

    CROSS-MODEL SCORE NORMALIZATION
    4.
    发明申请

    公开(公告)号:US20250131337A1

    公开(公告)日:2025-04-24

    申请号:US19005559

    申请日:2024-12-30

    Applicant: Dropbox, Inc.

    Inventor: Jongmin Baek

    Abstract: Computer-implemented techniques encompass using distinct machine learning sub-models to score respective types of candidate content for the purpose of providing personalized content suggestions to end-users of a content management system. The relevancy scores generated by the distinct sub-models are mapped to expected end-user interaction scores of the candidate content scored. Content suggestions are provided at end-users' computing devices where the suggested content is selected from the candidate content based on the expected end-user interaction scores of the candidate content. For each distinct sub-model, a normalizing mapping function is solved using an optimizer that maps the relevancy scores generated by the sub-model for the candidate content to expected end-user interaction scores for the candidate content. The expected end-user interaction scores are comparable across the distinct sub-models and can be used to rank content suggestions across the distinct sub-models.

    FILTER FOR CAMERA UPLOAD
    5.
    发明申请

    公开(公告)号:US20250130900A1

    公开(公告)日:2025-04-24

    申请号:US18382811

    申请日:2023-10-23

    Applicant: Dropbox, Inc.

    Inventor: Qimeng Li

    Abstract: A system and method are provided for automatically selecting a subset of photos to be backed up by an intelligent photo upload service provided by a content management system. The intelligent photo upload service is provided to a client application on a client device that uses one or more machine-learning models to determine which photos have the one or more attributes. The photos that have been determined to have the one or more attributes may be uploaded. Once uploaded the one or more attributes may be used to label the respective photos with metadata indicating the presence of the one or more attributes. The one or more attributes may be used to filter photos at the content management system.

    GENERATING FILLABLE DOCUMENTS AND FILLABLE TEMPLATES IN A COLLABORATIVE ENVIRONMENT

    公开(公告)号:US20250045519A1

    公开(公告)日:2025-02-06

    申请号:US18920682

    申请日:2024-10-18

    Applicant: Dropbox, Inc.

    Abstract: A collaborative content management system (CMS) is disclosed herein for generating templates for received documents. The disclosed CMS recognizes that a document selected by a user for processing was previously processed by the CMS and that a user has previously added particular overlaid fillable fields to the document. When the determination is made, the system generates a recommendation to create a template of the document with the previously added overlaid fillable fields. In some embodiments, the CMS makes the recommendation to generate a template when the user creates, in a received document, identical overlaid fillable fields or field types to those created in the previously processed document.

    Cross-model score normalization
    7.
    发明授权

    公开(公告)号:US12210948B2

    公开(公告)日:2025-01-28

    申请号:US18659482

    申请日:2024-05-09

    Applicant: Dropbox, Inc.

    Inventor: Jongmin Baek

    Abstract: Computer-implemented techniques encompass using distinct machine learning sub-models to score respective types of candidate content for the purpose of providing personalized content suggestions to end-users of a content management system. The relevancy scores generated by the distinct sub-models are mapped to expected end-user interaction scores of the candidate content scored. Content suggestions are provided at end-users' computing devices where the suggested content is selected from the candidate content based on the expected end-user interaction scores of the candidate content. For each distinct sub-model, a normalizing mapping function is solved using an optimizer that maps the relevancy scores generated by the sub-model for the candidate content to expected end-user interaction scores for the candidate content. The expected end-user interaction scores are comparable across the distinct sub-models and can be used to rank content suggestions across the distinct sub-models.

    Personalized spelling correction
    8.
    发明授权

    公开(公告)号:US12210518B2

    公开(公告)日:2025-01-28

    申请号:US16871398

    申请日:2020-05-11

    Applicant: Dropbox, Inc.

    Inventor: Matt Barta

    Abstract: A content management system generates a personalized vocabulary for a user by analyzing content items associated with a namespace of the user. The content management system receives a query. The content management system determines that the at least one search term does not occur at least a threshold number of times in the personalized vocabulary. The content management system identifies a prefix of the at least one search term. The content management system parses the personalized vocabulary based on the prefix to identify a set of search term options. The content management system generates a personalized ranking of the search term options based on (a) an edit distance; (b) a frequency of each search term option's occurrence in the personalized vocabulary; or (c) a combination of (a) and (b). The content management system presents at least a top ranked search term option to the user.

    Identifying related user accounts based on authentication data

    公开(公告)号:US12177199B2

    公开(公告)日:2024-12-24

    申请号:US17528819

    申请日:2021-11-17

    Applicant: Dropbox, Inc.

    Inventor: Anton Mityagin

    Abstract: In some embodiments, upon detecting malicious activity associated with a user account, a content management system can identify other user accounts related to the malicious user account. The content management system can identify related user accounts by comparing authentication information collected for the malicious user account with authentication information collected for other user accounts. Authentication information can include IP address information, geographic information, device type, browser type, email addresses, and/or referral information, for example. The content management system can compare the content items associated with the malicious user account to content items associated with other user accounts to determine relatedness or maliciousness. After identifying related malicious user accounts, the content management system can block all related malicious user accounts.

Patent Agency Ranking