Vector-Graph Embedding Ensemble for Complex Query Augmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing QA systems struggle to accurately handle complex or ambiguous queries, often failing to identify relevant information and understand the true intent behind user queries, leading to inaccurate and unsatisfactory responses.
Innovation Solution
Implementing an ensemble of vector and graph-based embeddings that encode knowledge bases using two pipelines: semantic vector augmentation and knowledge graph augmentation, which generate vector embeddings and knowledge triplets from unstructured and structured data, respectively, to augment user queries and enhance the accuracy of large language models (LLMs).
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional QA systems are used to handle queries, then the system structure remains simple, but the accuracy and ability to understand complex or ambiguous queries deteriorates
Solution Approach 1:
The patent segments the query processing system into multiple specialized components: a knowledge graph processor that handles structured relationships, a vector database processor that handles semantic similarity, and a traditional search processor. Each component specializes in specific types of query understanding, allowing the system to achieve high accuracy on complex queries while maintaining modular architecture that manages complexity through division of labor.
2Productivity
If an ensemble of vector and graph based embeddings is implemented to augment user queries, then the accuracy and efficiency of QA systems improves, but the processing overhead and computational resources required increases
Solution Approach 1:
The system performs preliminary action by pre-processing documents during ingestion to create both knowledge graph triplets and vector embeddings in advance. This pre-computation stores semantic relationships and meanings beforehand, so that during query processing, the system can quickly retrieve and combine these pre-computed representations without performing heavy computation in real-time, thus reducing processing overhead while maintaining high accuracy.
3Loss of information
If multiple pipelines (semantic vector augmentation and knowledge graph augmentation) are used to encode knowledge bases, then the comprehensiveness of information retrieval improves, but the system complexity and processing time increases
Solution Approach 1:
The patent merges multiple information representation pipelines by combining knowledge graph triplets (which capture structured relationships) with vector embeddings (which capture semantic meaning) into a unified query augmentation process. Both pipelines process documents in parallel during ingestion, and their outputs are integrated during query processing to provide comprehensive information retrieval. This merging approach ensures complete information capture while managing complexity through parallel processing and unified architecture.
Data Source
AI summary
Methods, systems, apparatuses, devices, and computer program products are described. A system may obtain a set of documents for input into a query response system, generate a set of vector embeddings based on the set of documents and a semantic vector augmentation pipeline, and generate a set of knowledge graphs based on the set of documents and a knowledge graph augmentation pipeline, where each knowledge graph includes a set of multiple knowledge graph triplets. The system may receive a user query and augment the user query to generate an augmented prompt using at least one or more vector embeddings from the set of vector embeddings and one or more knowledge graph triplets from the set of knowledge graphs. The system may provide, to a large language model (LLM), the augmented prompt as an input and may receive, as an output of the LLM, a response to the augmented prompt.