Web Page Text Classification via Image Proximity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in effectively extracting information about images from web pages, as text related to images is often difficult to identify and classify, impacting indexing and search functionalities.
Innovation Solution
A system that classifies text based on the proximity of text to images within web pages, using a trained classifier with both positive and negative examples to distinguish relevant from irrelevant text, allowing for the identification and description of images without analyzing the images themselves.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If text from web pages is used for image indexing, then image information can be extracted, but it is difficult to determine which text is actually related to the images
Solution Approach 1:
The patent segments text into different regions based on spatial proximity to images. Text is divided into near-text (within threshold distance) and far-text (beyond threshold distance), allowing the system to selectively process only the relevant segments for image indexing while excluding irrelevant portions.
Solution Approach 2:
The patent applies different quality criteria to different text regions. Near-text regions are assigned higher relevance weights and undergo more rigorous classification, while far-text regions are either excluded or given minimal weight, ensuring that local text quality varies according to its spatial relationship with the image.
2Quantity of substance
If all text in web pages is used for classification, then more information is available, but irrelevant text reduces classification accuracy
Solution Approach 1:
The patent extracts only the relevant portion of text (near-text) that is spatially close to the image, separating it from the bulk of irrelevant text (far-text). This extraction process maintains sufficient text volume for meaningful classification while eliminating noise that would degrade accuracy.
Solution Approach 2:
The patent applies partial action by using only a subset of available text (those within threshold distance) rather than all text. This selective approach provides just enough text information for accurate classification without the detrimental effect of including excessive irrelevant text.
3Measurement precision
If a classifier is trained with both positive and negative examples, then classification accuracy improves, but the training process becomes more complex
Solution Approach 1:
The patent performs preliminary action by pre-segmenting text into near-text and far-text categories before training the classifier. This preliminary organization creates naturally occurring positive examples (near-text) and negative examples (far-text), simplifying the training process while maintaining high classification accuracy.
Solution Approach 2:
The patent introduces spatial threshold distance as an intermediary criterion that automatically generates balanced training examples. This intermediary mechanism objectively separates relevant from irrelevant text, providing a structured approach to creating training data without manual intervention.
4Ease of operation
If text proximity to images is used to determine relevance, then relevant text can be identified, but text from large distances may still contain useful information
Solution Approach 1:
The patent uses parameter changes by adjusting the threshold distance parameter to control the balance between identification ease and information retention. By tuning this parameter, the system can expand or contract the near-text region to include or exclude distant text based on specific application requirements.
Solution Approach 2:
The patent applies dynamics by making the text selection criterion flexible rather than fixed. The threshold distance can be dynamically adjusted based on image size, document layout, and application needs, allowing the system to adaptively include or exclude distant text while maintaining operational simplicity.
Data Source
AI summary
Text in web pages or other text documents may be classified based on the images or other objects within the webpage. A system for identifying and classifying text related to an object may identify one or more web pages containing the image or similar images, determine topics from the text of the document, and develop a set of training phrases for a classifier. The classifier may be trained and then used to analyze the text in the documents. The training set may include both positive examples and negative examples of text taken from the set of documents. A positive example may include captions or other elements directly associated with the object, while negative examples may include text taken from the documents, but from a large distance from the object. In some cases, the system may iterate on the classification process to refine the results.