Semantic Prompt Retrieval Using Vector Search Metadata
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing ML models are constrained by their reliance on structured data and struggle to effectively utilize unstructured data, such as knowledge articles, chat transcripts, and emails, which lack context and are often irrelevant when using traditional keyword searching.
Innovation Solution
A search retriever module that processes unstructured data through a vector search operation, generating a search retriever object with metadata to perform context-dependent searches, integrated with a prompt generator to create prompts for ML models, allowing them to generate relevant responses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional keyword searching is used on unstructured data, then the search process is simple and fast, but the search results lack context and are often irrelevant
Solution Approach 1:
The patent introduces an intermediary processing layer that converts unstructured data into structured representations with contextual metadata. This intermediary step bridges the gap between simple keyword search and complex semantic understanding, enabling relevant search results while maintaining operational simplicity through automated context extraction and structuring.
Solution Approach 2:
The system performs preliminary processing of unstructured data before the actual search operation, extracting and organizing contextual information in advance. This preliminary action prepares the data in a structured format that enables precise searching without requiring complex processing during the search execution phase.
2Quantity of substance
If unstructured data is incorporated into ML model prompts, then the model can access more information, but the data compatibility with model algorithms decreases
Solution Approach 1:
The patent transforms unstructured data by changing its structural parameters into formats compatible with ML model algorithms. This includes organizing data into standardized fields, extracting relevant features, and formatting contextual information in ways that align with model input requirements, thereby maintaining both data volume and compatibility.
Solution Approach 2:
The system segments unstructured data into distinct, manageable components that can be individually processed and integrated into model prompts. By dividing complex unstructured data into structured segments with clear metadata, the system enables effective utilization of large data volumes while ensuring each segment maintains algorithmic compatibility.
3Measurement precision
If context-dependent semantic search is performed on unstructured data, then search result accuracy improves, but the processing time and computational resources increase
Solution Approach 1:
The system performs preliminary extraction and structuring of contextual information from unstructured data before the search operation. This advance preparation creates ready-to-use structured representations that enable fast, accurate searching without requiring intensive computational resources during the actual search execution.
Solution Approach 2:
The patent extracts only the essential contextual information needed for accurate searching from the full unstructured data set. By selectively extracting relevant contextual elements rather than processing entire documents, the system achieves high search accuracy while minimizing processing time and computational overhead.
Data Source
AI summary
Disclosed herein are system, method, and computer program product aspects for semantic search in a model-based prompt builder system. A system generates a search retriever object based on a search index comprising unstructured data. The search retriever object includes metadata specifying one or more details of a vector search operation to be performed on the search index. The system obtains search results by performing the vector search on the search index based on the one or more details of the vector search operation provided by the search retriever object and a search query. The system provides the search results to a prompt generator configured to use a model to generate a reply to a prompt request requiring the search results.