Worked on the Unstructured-IO/unstructured repository to address a specific issue in ontology image categorization within HTML content. Focused on improving document processing and HTML parsing by ensuring that images located inside div or span elements without accompanying text are correctly annotated as images in the ontology. Utilized Python to implement the fix, which prevents misclassification and enhances the accuracy of downstream data extraction and ontology alignment. Developed targeted tests to cover scenarios involving empty-text containers, providing regression safety. This work contributed to higher data quality for image annotations in HTML-derived content, leveraging skills in ontology mapping and document processing.