Chat-Guided Search Interface for Real-Time Attribute Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional keyword-based search methods fail to capture user intent and semantic relationships, leading to poor search performance in high-dimensional embedding spaces, especially with short text inputs.
Innovation Solution
Utilize a large language model (LLM) to enhance user input by generating complementary data, such as additional keywords or synonyms, and transform it into vector embeddings for improved search performance, incorporating user-specific context and preferences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional keyword-based search methods are used, then the search system is simple and easy to implement, but the search accuracy and ability to capture user intent deteriorates
Solution Approach 1:
The patent introduces an LLM as an intermediary component between the user's short text input and the vector search system. The LLM expands the brief query into more comprehensive search terms and contextual information, enabling accurate vector embeddings to be generated without requiring the user to provide detailed inputs. This mediator resolves the contradiction by enhancing search accuracy while keeping the user interface simple.
Solution Approach 2:
The system performs preliminary processing of the user's short text input by using the LLM to generate expanded search terms and contextual information before creating vector embeddings. This preliminary expansion action ensures that even brief queries are transformed into comprehensive search representations, improving search accuracy without increasing the apparent complexity for the user.
2Ease of operation
If short text inputs are used for search, then the ease of operation is improved, but the search performance and relevance of results deteriorates
Solution Approach 1:
The LLM acts as a mediator that takes the user's simple short text input and transforms it into enriched search queries with expanded keywords, synonyms, and contextual information. This allows users to benefit from ease of operation while the system achieves reliable search performance through the LLM's generative enhancement of the query.
Solution Approach 2:
The system changes the parameter of the search query from a brief text string to an expanded set of search terms and contextual information generated by the LLM. This parameter transformation maintains the ease of operation for users providing short inputs while significantly improving search performance through the enriched query representation.
3Measurement precision
If LLM is used to enhance user input, then the search accuracy and relevance is improved, but the processing time and computational resources increase
Solution Approach 1:
The system applies partial action by using the LLM to generate only the most essential expanded search terms and contextual information needed for accurate vector search, rather than generating exhaustive content. This selective enhancement improves search accuracy while limiting the time and computational resources consumed by the LLM processing.
Data Source
AI summary
A computer-implemented method is disclosed. The method includes: presenting, via a first user interface, an initial set of user interface elements associated with results of a search query; generating user prompt data for soliciting user selection of a subset of object attributes associated with one or more of the search query results; presenting the user prompt data as chat outputs in a second user interface; receiving, via the second user interface, an indication of one or more preferred object attributes; and updating the first user interface in real-time to display user interface elements corresponding to a reduced set of searchable objects associated with the one or more preferred object attributes.