Embedding Object Memory with Vector Storage for Semantic Retrieval
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computing systems face inefficiencies in storing and retrieving large quantities of content information due to reliance on keyword searches that do not consider abstract meanings, leading to inaccuracies and high memory requirements.
Innovation Solution
Utilizing semantic embedding models to generate embeddings that are stored in an embedding object memory, allowing for semantic proximity matching and reducing memory footprint through vector representations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If keyword searches are used to iterate through content objects, then the system can find related content objects, but the search accuracy is reduced because abstract meanings are not considered
Solution Approach 1:
The patent transforms the search mechanism from keyword-based text matching to vector-based semantic similarity comparison. Content objects are converted into numerical embedding vectors that capture their abstract meanings, enabling the system to measure semantic proximity quantitatively through vector operations such as cosine similarity, thereby resolving the contradiction between search accuracy and mechanism complexity
Solution Approach 2:
The patent replaces the mechanical keyword-matching system with a semantic embedding-based vector comparison system. Instead of iterating through text strings and comparing keywords, the system uses pre-computed embedding vectors and mathematical similarity measures, substituting a complex text-processing mechanical system with a more efficient mathematical computation approach
2Quantity of substance
If large quantities of content information are stored for searching, then more content objects are available for retrieval, but the memory requirements increase significantly
Solution Approach 1:
The patent extracts only the essential semantic features of content objects into compact embedding vectors, storing these vectors instead of the full content objects. This extraction process retains the meaningful information needed for similarity search while discarding redundant data, thereby reducing memory requirements while preserving the ability to retrieve relevant content
Solution Approach 2:
The patent creates compressed numerical representations (embeddings) as copies of the semantic essence of content objects. These embedding vectors serve as lightweight proxies that capture the meaningful characteristics of the original content without requiring storage of the complete content objects, enabling efficient memory usage while maintaining search capability
3Productivity
If keyword searches are used to process large quantities of content information, then the system can operate with simple mechanisms, but computational efficiency is reduced
Solution Approach 1:
The patent applies preliminary action by pre-computing embedding vectors for content objects and storing them in advance. When a search query arrives, the system only needs to compute the embedding for the query and compare it against the pre-computed embeddings, avoiding the need to process and analyze the full text of each content object during the search operation, thereby significantly improving computational efficiency
Data Source
AI summary
Methods, systems, and media for storing entries in and/or retrieving information from an embedding object memory are provided. In some examples, a content item is received that has content data. The content data associated with the content item may be provided to one or more semantic embedding models that generate semantic embeddings. From one or more of the semantic embedding models, one or semantic embeddings may be received. The one or more semantic embeddings may then be inserted into the embedding object memory. The semantic embeddings may be associated with respective indications corresponding to a reference to source data associated with the semantic embeddings. Further, the insertion may trigger a spatial storage operation to store a vector representation of the one or more semantic embeddings. A plurality of collections of stored embeddings may be received from the embedding object memory, based on a provided input, to determine an action.