Dynamic Labeling of Functionally Equivalent DOM Nodes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Identifying objects with similar functionality on web pages is challenging due to the large number of objects, leading to inaccuracies in machine learning algorithms if objects with the same functionality are not included in the training set.
Innovation Solution
A system that transforms web pages into Document Object Model (DOM) trees, recursively identifies and labels functionally equivalent nodes, and uses these labeled nodes to train machine learning algorithms, ensuring more accurate recognition of object functions without human intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning algorithms are trained on web page objects, then object function identification is performed, but accuracy decreases when objects with the same functionality are not included in the training set
Solution Approach 1:
The patent merges functionally equivalent objects by identifying them through DOM tree traversal and request comparison. When multiple objects (e.g., buttons, links) produce the same request when interacted with, they are grouped together and labeled as functionally equivalent, ensuring all instances of a function are captured in the training set.
Solution Approach 2:
The system performs self-labeling by automatically comparing requests generated from object interactions. The machine learning system itself generates the training labels by observing which objects produce identical requests, eliminating the need for manual annotation while ensuring comprehensive coverage of functional equivalents.
2Measurement precision
If manual labeling of web page objects is performed, then training data is created, but time consumption increases due to the large number of objects
Solution Approach 1:
The system automatically labels objects by having the machine learning algorithm itself generate the labels through request comparison. The system traverses the DOM tree, interacts with objects, compares their requests, and assigns labels autonomously, eliminating manual labeling time while maintaining high training data quality.
Solution Approach 2:
The system performs preliminary DOM tree traversal and request generation before final labeling. By pre-interacting with objects and capturing their requests in advance, the system prepares all necessary labeling information beforehand, significantly reducing the time required for the actual labeling process.
3Quantity of substance
If all objects in a web page are analyzed for functional equivalence, then training set completeness improves, but processing complexity increases
Solution Approach 1:
The patent segments the web page into a hierarchical DOM tree structure, processing objects level by level from root to leaves. This segmentation allows the system to manage complexity systematically, traversing and comparing objects in organized groups rather than processing all objects simultaneously, while still ensuring complete functional equivalence analysis.
Data Source
AI summary
A baseline request produced from an annotated node of a document object model (DOM) tree and a label assigned to the annotated node are obtained. The label is assigned to a set of neighboring nodes of the DOM that perform a same function by recursively causing the system to, for each neighboring node to the annotated node in the DOM tree an additional request produced in response to performance of simulated human interaction with the neighboring node is identified, if the additional request matches the baseline request, the label is assigned to the neighboring node, and the neighboring node is selected to be the annotated node.