Interest Prediction Model for Content Taxonomy Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Online services face challenges in accurately mapping user-posted content, especially images, to relevant interest nodes in a taxonomy, as the textual content often does not exactly match the text labels of the interest nodes.
Innovation Solution
The solution involves training an interest prediction model using various feature sets, including text-based, co-occurrence, TF/IDF, and embedding features, to identify candidate interest nodes for user-posted content. This model aggregates feature sets to predict the most likely interest nodes associated with the content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional text-matching methods are used to map user content to interest nodes, then the system is simple and fast, but the mapping accuracy deteriorates because user text does not exactly match taxonomy labels
Solution Approach 1:
The patent introduces an intermediary interest prediction model that acts as a mediator between user-generated content and the interest taxonomy. Instead of directly matching user text to taxonomy labels, the system uses the model to predict interest nodes based on multiple feature sets including text features, co-occurrence features, and embedding features. This intermediary layer resolves the contradiction by enabling accurate mapping without requiring exact text matches.
Solution Approach 2:
The patent combines multiple different feature sets (text-based features, co-occurrence features, TF/IDF features, and embedding features) into a composite feature representation. This composite approach allows the system to capture various aspects of user content simultaneously, improving mapping accuracy by considering multiple dimensions of content meaning rather than relying on a single feature type.
2Measurement precision
If multiple feature sets are aggregated to improve mapping accuracy, then the precision improves, but the computational complexity and processing time increase
Solution Approach 1:
The patent performs preliminary actions by pre-computing and storing embedding features for user content and organizing the interest taxonomy in advance. The embedding model pre-processes user content to generate vector representations that capture semantic meaning. This preliminary processing reduces the computational burden during actual mapping operations, as the system can directly compare pre-computed embeddings with taxonomy node representations rather than performing complex analysis in real-time.
3Adaptability or versatility
If the interest taxonomy is comprehensive with many nodes, then the coverage and adaptability improve, but the difficulty of correctly mapping content to the appropriate node increases
Solution Approach 1:
The patent applies local quality by generating different types of features that capture different aspects of content-interest relationships. Each feature set (text features, co-occurrence features, embedding features) provides a different local perspective on the content. The system aggregates these diverse local views to make a comprehensive mapping decision, allowing the model to leverage the strengths of each feature type for different aspects of the mapping task.
Data Source
AI summary
System and methods are presented for associating a user-posted content item with an interest node of an interest taxonomy. A corpus of content items and an interest taxonomy are maintained. The interest taxonomy comprises interest nodes organized in a hierarchical organization, each node having a text label descriptive of the interest node. Additionally, the content items of the corpus are associated with one or more interest nodes of the interest taxonomy. Upon receiving a user-posted content item, feature sets of the received content item are generated, these feature sets based on features and/or aspects of the received content item. After generating at least one feature set, the at least one feature set is provided to an interest prediction model that generates candidate interest nodes for the user-posted content item. At least some of the candidate interest nodes are associated with the user-posted content item in the corpus.