Cross-lingual Listing Pairing via Neural Feature Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems fail to effectively pair similar product listings across different languages, neglecting the consideration of image and text characteristics, which hampers the accuracy of machine translation systems.
Innovation Solution
A Listing Engine that utilizes deep-learned neural network models for feature extraction from listing images and text, calculating similarity scores to pair listings in different languages based on their likelihood of describing the same item, and inserts these pairs into a machine translation system's training data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional systems are used to pair product listings across different languages, then the system simplicity is maintained, but the pairing accuracy deteriorates due to neglecting image and text characteristics
Solution Approach 1:
The system segments the product listing pairing process into multiple independent analysis components: image feature extraction, text feature extraction, and similarity scoring. Each component processes specific characteristics separately and contributes to the overall pairing decision, enabling comprehensive analysis without overwhelming system complexity
Solution Approach 2:
The patent introduces neural network models as intermediary components that bridge the gap between raw listing data and pairing decisions. These models extract meaningful features from images and text, transforming unstructured data into comparable representations that improve pairing accuracy while managing complexity through specialized processing layers
2Measurement precision
If image and text features are extracted using deep-learned neural network models, then the feature extraction accuracy is improved, but the computational complexity increases
Solution Approach 1:
The system performs preliminary feature extraction using pre-trained neural network models before the actual pairing decision. Image and text features are extracted in advance and stored as processed data, reducing the computational burden during real-time pairing operations and enabling accurate comparisons without excessive complexity during execution
Solution Approach 2:
The patent uses neural network models to create simplified representations (copies) of the original complex image and text data. These feature vectors serve as compressed versions that capture essential characteristics while reducing dimensionality, enabling accurate similarity comparisons with lower computational requirements
3Reliability
If similarity scores are calculated based on multiple features, then the pairing reliability is improved, but the processing time increases
Solution Approach 1:
The system calculates similarity scores based on multiple features but applies selective processing: essential features (image similarity, text similarity) are fully evaluated, while secondary features are processed with reduced depth or weighted differently. This approach maintains reliable pairing decisions while reducing overall processing time by avoiding exhaustive analysis of all possible features
Data Source
AI summary
In various example embodiments, a system and method for a Listing Engine that translates a first listing from a first language to a second language. The first listing includes an image(s) of a first item. The Listing Engine provides as input to an encoded neural network model a portion(s) of a translated first listing and a portions(s) of a second listing in the second language. The second listing includes an image(s) of a second item. The Listing Engine receives from the encoded neural network model a first feature vector for the translated first listing and a second feature vector for the second listing. The first and the second feature vectors both include at least one type of image signature feature and at least one type of listing text-based feature. Based on a similarity score of the first and second feature vectors at least meeting a similarity score threshold, the Listing Engine generates a pairing of the first listing in the first language with the second listing in the second language for inclusion in training data of a machine translation system.