Personal Entity Modeling on Mobile Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing mobile device technologies lack the ability to effectively model and utilize personal entities, which are unique to individual users, leading to a suboptimal user experience due to the reliance on public knowledge bases that often miss personal interactions and interests.
Innovation Solution
A system that detects and models personal entities in an embedding space, using continuous bag of words-like training models to predict personal entities based on context and interactions, allowing for personalized recommendations and enhancements in search and advertising.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If public knowledge bases are used to model entities, then the system can leverage existing large-scale entity data, but it fails to capture personal entities unique to individual users
Solution Approach 1:
The system segments the knowledge base into two distinct components: a public knowledge base for general entities and a personal knowledge base for user-specific entities. This segmentation allows the system to handle personal entities separately, capturing user-specific information that would be lost in a unified public knowledge base approach.
Solution Approach 2:
The patent introduces an intermediary mechanism - the personal entity detector and embedding generator - that bridges the public knowledge base and personal user data. This intermediary processes user interactions and content to extract personal entities, converting them into embeddings that can be stored and utilized in the personal knowledge base.
2Productivity
If the system tracks and models personal entities from user content, then personalized predictions can be generated, but the complexity of entity detection and modeling increases
Solution Approach 1:
The patent replaces complex rule-based entity detection mechanisms with a machine learning-based embedding system. By using continuous bag of words (CBOW) models and neural network embeddings, the system automatically learns entity representations from data, eliminating the need for manual entity detection rules and reducing overall system complexity while improving personalization accuracy.
Solution Approach 2:
The system transforms personal entities from discrete categorical data into continuous embedding vectors. This parameter transformation allows the system to represent entities in a dense vector space where similarity and relationships can be computed efficiently, simplifying the modeling process while capturing nuanced personal entity characteristics.
3Measurement precision
If embeddings are continuously learned and adjusted over time, then the personal knowledge base becomes more accurate, but computational resources are consumed
Solution Approach 1:
The system implements periodic training of embedding models rather than continuous real-time updates. By training embeddings at scheduled intervals or based on accumulation of sufficient user interaction data, the system achieves accurate personal entity representations while avoiding the continuous computational overhead of real-time embedding updates.
Solution Approach 2:
The patent performs preliminary embedding generation and entity detection during periods of lower system load or in batch processing mode. By pre-computing embeddings and storing them in the personal knowledge base, the system reduces the computational burden during active user interactions, consuming energy more efficiently while maintaining high measurement precision.
Data Source
AI summary
Systems and methods are provided for a personal entity modeling for computing devices. For example, a computing device comprises at least one processor and memory storing instructions that, when executed by the at least one processor, cause the mobile device to perform operations including identifying a personal entity in content generated for display on the mobile device, generating training examples for the personal entity from the content, and updating an embedding used to model the personal entity using the training examples. The embedding may be used to make predictions regarding the personal entity. For example, the operations may also include predicting an association between a first personal entity displayed on the computing device and a second entity based on the embedding, and providing a recommendation, to be displayed on the computing device, related to the second entity.