Machine-Learning Query Attribute Features for Relevant Product Retrieval
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing online search systems often return irrelevant results due to partial matches, and limiting results to only relevant items may exclude potentially relevant results, leading to inefficiencies in product retrieval.
Innovation Solution
An online system generates product attribute features using machine learning models to identify tags associated with queries, ranks these tags based on historical conversion rates, and selects candidate products for display based on generated scores, ensuring relevance and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the online system searches the item catalog using provided search terms to identify relevant items, then the search results include items estimated to be relevant to the search query, but the results often include irrelevant items from partial matches
Solution Approach 1:
The patent segments the search result filtering process into multiple stages: first retrieving items based on search terms, then applying machine learning models to generate relevance scores, and finally filtering/sorting results based on these scores. This multi-stage segmentation allows the system to maintain a comprehensive result set while improving relevance through progressive filtering.
Solution Approach 2:
The patent introduces machine learning models as an intermediary between the search query and final results. These models generate relevance scores that act as a mediator to evaluate and rank items, enabling the system to distinguish relevant from irrelevant results more effectively than simple keyword matching alone.
2Reliability
If the online system limits search results to only actually relevant results, then the relevance of results is improved, but potentially relevant results may be excluded
Solution Approach 1:
The patent implements dynamic relevance scoring using machine learning models that adapt to different search queries and user contexts. The system generates relevance scores dynamically rather than using fixed filtering rules, allowing it to adjust the threshold for what constitutes a relevant result based on the specific query and historical data, thus balancing relevance with coverage.
Solution Approach 2:
The patent changes the parameter of result selection from binary (relevant/not relevant) to continuous (relevance score). By using machine learning models to generate nuanced relevance scores, the system can adjust selection thresholds to balance between including only highly relevant results and capturing potentially relevant results, thereby adapting to different search scenarios.
3Measurement precision
If the online system uses machine learning models to generate product attribute features and scores, then the relevance and accuracy of product retrieval is enhanced, but the system complexity increases
Solution Approach 1:
The patent segments the complex machine learning pipeline into distinct components: a first model for generating product attribute features from search queries, and a second model for scoring candidate products. This segmentation allows each model to specialize in a specific task, improving overall accuracy while making the system more manageable and potentially enabling parallel processing to mitigate complexity concerns.
Solution Approach 2:
The patent introduces machine learning models as intermediary processing layers between the search interface and the product catalog. These models act as intelligent mediators that translate user queries into structured features and then evaluate products based on those features, adding measurement precision without requiring direct complex interactions between all system components.
Data Source
AI summary
An online system receives a query from a user. An online system generates a prompt to provide to a first machine learning model to determine tags associated with the query. An online system provides the prompt to the first machine learning model. An online system receives as output a set of query tags associated with the query. An online system obtains a list of ranked product tags associated with the query, wherein the product tag is ranked according to a conversion rate of products matching the product tag when users submitted a historical search query. An online system identifies a set of candidate products for the search query. An online system, for each candidate product, provides the set of features to a second machine-learning model to generate a score for the candidate and selects at least a subset of the candidate products based on the generated scores for the candidate products.