Statistical Model for Content Relevance Ranking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for ranking parameters of content items in advertising systems often result in unsatisfactory advertisement placement due to flawed assumptions and methodologies, leading to less-relevant advertisements being delivered to users, which affects click-through rates and conversion rates.
Innovation Solution
A computer-implemented method and system that utilize a statistical model, trained using historical data, to estimate revised relevance scores of parameters by considering co-occurrences and pruning initial relevance scores based on co-occurrence frequencies, allowing for more accurate ranking of parameters and improved advertisement placement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional methods are used to rank parameters of content items, then the ranking process is simple and fast, but the advertisement placement becomes less relevant and click-through rates decrease
Solution Approach 1:
The patent segments the relevance scoring process into multiple independent parameter rankings (e.g., keyword relevance, semantic cluster relevance, category relevance, user visit history relevance) that are calculated separately and then combined. This allows each parameter to be scored independently using simple methods while the overall system achieves higher precision through the aggregation of multiple segmented rankings.
Solution Approach 2:
The patent introduces an intermediary statistical model that takes multiple initial relevance scores as inputs and produces a revised relevance score. This intermediary component (the statistical model functioning as a mediator) combines multiple simple rankings into a more accurate composite ranking, resolving the contradiction between simplicity and precision.
2Reliability
If multiple parameter sources are combined to obtain initial relevance scores, then the scoring becomes more comprehensive, but the processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary actions by pre-calculating and storing parameter co-occurrence frequencies and statistics before the actual ranking process. This preliminary preparation allows the statistical model to quickly compute revised relevance scores during runtime without performing complex real-time analysis of all parameter combinations, thus maintaining high reliability while reducing processing time.
Solution Approach 2:
The patent extracts and utilizes only the essential co-occurrence statistics from the multiple parameter sources, rather than processing all possible parameter interactions. By taking out only the necessary co-occurrence data needed for the statistical model, the system achieves comprehensive scoring reliability without the full computational burden of analyzing all parameter combinations.
3Measurement precision
If parameter co-occurrences are analyzed and pruning is applied, then the model accuracy improves, but the computational complexity increases
Solution Approach 1:
The patent applies partial action by performing pruning only on parameters whose co-occurrence frequency falls below a predetermined threshold. Instead of analyzing all possible parameter combinations exhaustively, the system focuses computational resources only on the subset of parameters that have significant co-occurrence relationships, thereby improving precision while limiting the increase in computational complexity to only the necessary portion.
Data Source
AI summary
System and methods allow for ranking relevance of parameters of a content item. A method includes: receiving, using at least one processing circuit, a plurality of parameters of a content item and a plurality of corresponding initial relevance scores of the parameters indicating relevance of the parameters to the content item; estimating, using a statistical model, a plurality of revised relevance scores from the initial relevance scores, wherein each of the revised relevance scores is a function of at least two of the plurality of initial relevance scores; and ranking the plurality of parameters based on the revised relevance scores.