Multi-Channel Query Resolution Using Keyword and Embedding Scores
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional search query resolution techniques are limited by user knowledge gaps, leading to ineffective searches in complex domains like clinical settings, where users may not know the necessary keywords or specialties for their queries, resulting in null results.
Innovation Solution
A multi-modal and multi-channel approach that transforms search queries into keyword and embedding representations, measuring syntactic and semantic similarities across domain channels, and aggregates results using weighted similarity scores to enhance search resolution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional keyword searching techniques are used on limited search result features, then the search process is simple and fast, but the search effectiveness is poor when users lack domain knowledge
Solution Approach 1:
The patent introduces an intermediary system that includes a language model and embedding model. These intermediaries translate user queries into semantic representations and bridge the gap between user intent and domain-specific terminology, enabling effective search without requiring users to have specialized knowledge
Solution Approach 2:
The search process is segmented into multiple independent components: keyword representation generation, embedding representation generation, similarity score calculation for different channels, and result aggregation. This segmentation allows each component to specialize in specific tasks while maintaining overall system manageability
2Measurement precision
If users are required to fill in a lengthy form with specific domain knowledge to complete a search query, then the search can be more precise, but the ease of operation decreases
Solution Approach 1:
The system performs self-service by automatically generating keyword and embedding representations from the user's natural language query without requiring manual input of multiple fields. The language model and embedding model autonomously extract relevant features and compute similarity scores, eliminating the need for users to understand complex domain terminology
Solution Approach 2:
The system performs preliminary actions by pre-processing the user query through language modeling and embedding generation before the actual search. This preliminary transformation of the query into semantic representations enables precise matching against domain channels without requiring users to prepare detailed input forms
3Productivity
If basic keyword searching is performed on provider features, then the search process is straightforward, but it returns null results when there is a lack of keyword matches
Solution Approach 1:
The patent implements multi-channel searching that operates across multiple domain channels (e.g., provider specialties, conditions, procedures) simultaneously. Each channel is processed with both keyword and embedding representations, making the search system universally applicable to various query types and domain specifications without requiring separate search processes
Solution Approach 2:
The search technique uses a composite approach by combining multiple representation types (keyword representation and embedding representation) and multiple similarity score calculations. This composite methodology integrates different search strategies to produce comprehensive results that leverage both exact keyword matching and semantic similarity
4Loss of information
If multi-modal and multi-channel search techniques are implemented with keyword and embedding representations, then the semantic understanding of queries is deeper, but the computational complexity increases
Solution Approach 1:
The computational process is segmented into distinct stages: generating keyword representations, generating embedding representations, calculating keyword similarity scores for each channel, calculating embedding similarity scores for each channel, and aggregating results. This segmentation reduces the complexity of any single computational step while maintaining comprehensive semantic analysis
Data Source
AI summary
Various embodiments of the present disclosure provide query processing techniques for resolving queries in a complex search domain to improve upon traditional search resolutions within such domains. The techniques may include generating a keyword and an embedding representation for an agnostic search query. The keyword representation may be compared against source text attributes within one or more domain channels to generate a plurality of keyword similarity scores between the search query and features within a search domain. The embedding representation may be compared against source embedding attributes within the one or more domain channels to generate a plurality of embedding similarity scores between the search query and the features within the search domain. The keyword and embedding similarity scores may be aggregated to generate aggregated similarity scores for identifying an intermediate query resolution for the search query. The intermediate query resolution may be leveraged to resolve the query.