Natural Language Search Intent Classification for Vertical Query Routing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional online search systems lack the ability to perform natural language searches across different verticals, resulting in ambiguous or irrelevant results, requiring users to repeatedly reformulate keyword searches and incurring additional computational overhead.
Innovation Solution
The system employs machine learning models to identify search intentions in natural language queries and convert them into search parameters for specific verticals, combining results from multiple verticals into a single ranking, allowing users to perform efficient natural language searches across online systems without needing multiple queries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional keyword search systems are used, then search coverage across multiple verticals is achieved, but search result relevance and clarity deteriorate due to ambiguity
Solution Approach 1:
The search system is segmented into multiple independent vertical search modules (e.g., jobs, companies, schools, groups) rather than a single monolithic search. Each vertical module processes queries independently with vertical-specific parameters, allowing the system to maintain broad search coverage across multiple domains while ensuring precise, relevant results within each specific domain through dedicated processing logic.
2Measurement precision
If users perform multiple reformulated keyword searches to improve relevance, then search result accuracy improves, but computational overhead and processing time increase
Solution Approach 1:
The system performs preliminary classification of the user's natural language query to identify the intended vertical (e.g., determining whether the query seeks jobs, companies, or schools) before executing the search. This preliminary action eliminates the need for users to reformulate queries multiple times, as the system proactively routes the single query to the appropriate vertical module, thereby maintaining high accuracy while avoiding repeated computational overhead.
Solution Approach 2:
A natural language processing intermediary layer is introduced between the user query and the vertical search modules. This intermediary parses the natural language query, extracts search parameters, and routes the query to the appropriate vertical. This mediator translates ambiguous natural language into precise vertical-specific search parameters in one step, eliminating the need for multiple query reformulations and reducing computational waste.
3Measurement precision
If natural language processing with machine learning models is implemented, then search intention accuracy improves, but system complexity increases
Solution Approach 1:
The natural language processing system is segmented into multiple specialized machine learning models, with each model trained specifically for a particular vertical (e.g., one model for job-related queries, another for company queries). This segmentation allows each model to achieve high accuracy for its specific domain while keeping individual model complexity manageable. The system selects and applies only the relevant model based on the query type, avoiding the need for a single overly complex universal model.
Data Source
AI summary
The disclosed embodiments provide a system for performing a natural language search. During operation, the system applies a first machine learning model to a natural language query to predict one or more search intentions associated with the natural language query. Next, the system applies a second machine learning model to the natural language query to produce one or more search parameters associated with a first intention in the search intention(s), wherein the search parameter(s) include a field and a value of the field. The system then performs a first search of a first vertical associated with the first intention using the search parameter(s). Finally, the system generates a ranking containing a first set of search results from the first search of the first vertical and outputs at least a portion of the ranking in a response to the natural language query.