Item Ranking Model for High-Precision Catalog Search
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing search methods often retrieve multiple irrelevant items due to matching words in a product search field, leading to irrelevant results being displayed despite the target items being in the catalog, causing user frustration and inefficiency.
Innovation Solution
A system utilizing a trained machine learning model that scores and ranks item objects based on discriminative text analysis, including processes like tokenization, normalization, and category-based scoring, to ensure relevant results are prioritized.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If word-based matching is used to retrieve catalog products, then recall of potential items is improved, but precision of search results deteriorates due to inclusion of irrelevant items
Solution Approach 1:
The patent introduces an intermediary ranking system between the initial word-based retrieval and the final search results display. This ranking system uses multiple signals (engagement data, product signals, query signals) to evaluate and reorder the retrieved items, filtering out irrelevant ones before presentation to the user. The intermediary ranking process resolves the contradiction by maintaining high recall while achieving high precision through sophisticated scoring and re-ranking mechanisms.
2Adaptability or versatility
If comprehensive word matching is performed across all catalog products, then coverage of search results is improved, but computational resources deteriorate due to processing large datasets
Solution Approach 1:
The patent implements preliminary actions by pre-computing and storing engagement data, product signals, and query signals in structured formats before search execution. These pre-processed signals are organized in ways that enable efficient retrieval and scoring during actual search operations. This preliminary preparation reduces the computational burden during real-time searching while maintaining comprehensive coverage capabilities.
Solution Approach 2:
The patent changes parameters by transforming raw search data into normalized signal representations with different weights and priorities. By converting comprehensive product data into structured signals with associated importance weights, the system can efficiently process large datasets without exhaustive comparison, reducing computational resources while maintaining versatility through the flexible signal-weighting mechanism.
3Device complexity
If traditional ranking methods are used after retrieval, then simplicity of the system is maintained, but effectiveness of ranking deteriorates due to display of irrelevant results
Solution Approach 1:
The patent applies the composite materials principle by combining multiple types of signals (engagement data, product signals, query signals) into a composite ranking score. Each signal type contributes different characteristics to the final ranking evaluation, creating a robust and reliable ranking system that accurately distinguishes relevant from irrelevant results. This composite approach maintains systematic complexity while dramatically improving ranking effectiveness through multi-factor evaluation.
Data Source
AI summary
Examples relate to ranking item objects based on a query. An example includes receiving a query including a plurality of tokens, selecting at least one or more item objects including item object titles, ranking the one or more item objects using a ranking module, scoring each respective selected item object title and the query using a discriminative text module, receiving the scores for at least two categories for the one or more item object titles and the query, and applying a trained machine learning model to the query, the titles of each of the one or more item objects and the feature data for each respective item object of the one or more item objects, such that the trained machine learning model outputs an evaluation of the scores of the one or more item objects selected for the query.