Dynamic Feature Vector Generation for DOM Elements
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Feature engineering for machine learning algorithms is labor-intensive and time-consuming, especially when dealing with web pages, due to the complexity and non-standardization of web elements, making it difficult to train algorithms efficiently.
Innovation Solution
A system that automatically creates feature vectors for Document Object Model (DOM) tree elements in an unsupervised manner by tokenizing HTML strings, computing an information matrix, and deriving word vectors, which are then used to train machine learning models for classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual feature engineering is used for machine learning algorithms, then feature vectors can be created, but the process is labor-intensive and time-consuming
Solution Approach 1:
The system automatically generates feature vectors for DOM elements by tokenizing HTML strings, computing information matrices, and deriving word vectors without requiring manual human intervention. This self-service approach eliminates the need for human operators to manually create feature vectors, directly resolving the contradiction between training efficiency and time investment in feature engineering.
Solution Approach 2:
The patent replaces manual mechanical feature engineering processes with automated computational methods. Instead of humans manually analyzing and creating features, the system uses algorithms to tokenize HTML, compute information matrices, and generate word vectors, substituting human labor with automated mechanical/computational processes.
2Extent of automation
If manual feature engineering is performed, then features can be extracted, but it requires significant human labor and expertise
Solution Approach 1:
The feature engineering process is segmented into distinct automated steps: tokenization of HTML strings, computation of information matrices, and derivation of word vectors. This segmentation allows each complex task to be broken down into manageable automated operations, reducing the overall complexity perception while increasing automation extent.
Solution Approach 2:
The patent introduces intermediary computational structures (information matrices, word vectors, feature vectors) that mediate between the raw HTML data and the final machine learning inputs. These intermediaries automate the transformation process, reducing the need for human expertise while managing the complexity through structured computational steps.
3Adaptability or versatility
If standardization of web elements is attempted, then feature extraction becomes easier, but web pages are inherently non-standardized
Solution Approach 1:
The system creates a universal feature extraction approach that works across diverse, non-standardized web elements. By tokenizing HTML strings and computing information matrices that capture co-occurrence relationships, the system generates feature vectors that can represent any DOM element regardless of its specific HTML structure, achieving versatility without requiring element standardization.
Solution Approach 2:
The patent changes the parameters of feature extraction from structure-dependent to content-based. Instead of relying on standardized element types or structures, the system uses tokenization and information matrix computation to create features based on the actual content and co-occurrence patterns of elements, allowing adaptation to any HTML structure while maintaining ease of extraction.
Data Source
AI summary
A plurality of HyperText Markup Language (HTML) strings corresponding to a dataset of document object model (DOM) nodes is tokenized according to a tokenization scheme to produce a dictionary of tokens that occur in the dataset. A condensed dictionary of tokens in produced by removing low-value tokens from the vocabulary of tokens. An information matrix is computed based on the condensed dictionary of tokens, the information matrix being a set of values, a value of the set of values corresponding to a frequency of co-occurrence of a pair of tokens. A library of word vectors is derived from the information matrix. A feature vector of an HTML element is generated based at least in part on the library of word vectors, and a classification for the HTML element is obtained from a machine learning model as a result of inputting the feature vector into the machine learning model.