Product Name Extraction from Unstructured Web Pages
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to accurately extract product names from unstructured electronic documents, such as web pages, due to their lack of structured formats, making it difficult to identify relevant information like product names, organizations, people, or places.
Innovation Solution
An adaptive learning machine system is employed, utilizing a page recognition model and a name recognition model based on machine learning algorithms to identify product pages and filter terms within those pages, determining features like URL, structure, and content to accurately classify terms as product names or not.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If highly accurate rules are used to obtain document contents, then structured documents can be processed accurately, but unstructured web pages and documents become difficult to process
Solution Approach 1:
The patent transforms the approach from using fixed extraction rules to using machine learning models with adjustable parameters. The system learns optimal parameters from labeled training data, enabling accurate extraction from both structured and unstructured documents. The models adapt their parameters based on the specific document type and content patterns observed during training.
Solution Approach 2:
The patent replaces the mechanical rule-based system with an intelligent machine learning system. Instead of manually crafted rules that work only for structured documents, the system uses trained models that automatically adapt to different document formats including unstructured web pages, achieving both accuracy and versatility.
2Measurement precision
If machine learning models are trained on labeled data, then accurate identification of product names is achieved, but processing time and computational resources increase
Solution Approach 1:
The patent performs data labeling and model training in advance before actual product name extraction is needed. Labeled datasets are created beforehand, and machine learning models are trained offline. This preliminary preparation enables rapid, accurate extraction during actual use without time-consuming processing during the extraction phase itself.
Solution Approach 2:
The patent creates trained machine learning models as reusable copies that can be applied to multiple documents. Once a model is trained on labeled data, it becomes a persistent artifact that can be repeatedly used for extraction tasks without retraining, significantly reducing processing time for subsequent extractions while maintaining high accuracy.
Data Source
AI summary
Techniques for obtaining information from an electronic document include accessing a set of related electronic documents; identifying a product page associated with the set of related electronic documents using a page recognition model, the product page comprising a plurality of terms; filtering the plurality of terms into a first set of terms and a second set of terms, the first set of terms and the second set of terms including different terms of the plurality of terms, each term in the first set of terms identified as potentially being associated with a product name, and each term in the second set of terms identified as not being associated with a product name; and identifying each term in the first set of terms as being associated with a product name or not being associated with a product name with a name recognition model.