Webpage Transaction Extraction Using Synthetic Label Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing software applications require manually generated and updated scripts for each provider's website to extract transactions, which is not feasible due to the large number of providers, leading to inefficiency and complexity.
Innovation Solution
A machine learning model is trained using a small set of sample transactions to generate a large set of synthetic transactions with varied labels, enabling automatic transaction extraction from webpages with different structures and designs without the need for unique scripts for each provider.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manually generated and updated scripts are used for each provider's website, then transaction extraction can be performed accurately, but the complexity and effort increase significantly due to the large number of providers
Solution Approach 1:
The patent creates a universal script template that can extract transactions from multiple different provider websites. Instead of having separate scripts for each provider, a single template with parameterized selectors and label mappings can adapt to different website structures, reducing the number of scripts from thousands to a manageable set of templates
Solution Approach 2:
The patent uses parameter changes by allowing the script template to accept different configuration parameters for each provider (such as CSS selectors, label mappings, and XPath expressions). This enables the same base script to be customized for different providers through parameter adjustment rather than creating entirely new scripts
2Reliability
If manually generated and updated scripts are used for each provider's website, then transaction extraction can be performed, but the time and resources required for creation and updates become unsustainable
Solution Approach 1:
The patent performs preliminary action by pre-defining the script template structure, common selectors, and label mappings that can be reused across multiple providers. This preliminary setup reduces the time needed for each new provider integration, as the framework is already in place and only needs to be configured rather than created from scratch
Solution Approach 2:
The patent uses copying by replicating the universal script template across different providers with minimal modifications. Instead of manually creating unique scripts for each provider, the template is copied and adapted with provider-specific parameters, dramatically reducing the time and effort required for script deployment and updates
3Adaptability or versatility
If unique scripts are created for each provider, then specific website structures can be handled, but the system becomes difficult to maintain and update when website designs change
Solution Approach 1:
The universal script template maintains adaptability to different website structures through parameterized configurations while centralizing the maintenance point. When a website changes, updates are made to the template or its parameters rather than individual scripts, making the system easier to maintain
Solution Approach 2:
The patent merges the maintenance of multiple provider-specific scripts into a single universal template. This consolidation means that when website designs change, the maintenance effort is focused on one template rather than thousands of individual scripts, significantly improving ease of repair and updates
Data Source
AI summary
A method for training a machine learning model to automatically identify and extract transactions from webpages includes: obtaining sample data from a webpage, the sample data including: (i) a plurality of live transactions; and (ii) a first set of labels, each label in the first set of labels corresponding to a respective attribute of a plurality of different attributes of each of the plurality of live transactions; generating training data based on the sample data, the training data comprising: (i) a plurality of synthetic transactions; and (ii) a second set of labels including one or more labels that differ from each label included in the first set of labels; and training the machine learning model to automatically identify and extract transactions from webpages using the training data.