| from collections import Counter | |
| def quality(result, page, threshold=0.7): | |
| flags = [] | |
| if not page.features.get("blank") and page.features.get("ink_contrast", 1000) < 145: | |
| flags.append("low_contrast_or_blur") | |
| if not result.text.strip() and not page.features.get("blank"): | |
| flags.append("empty_nonblank") | |
| if 0 < len(result.text.strip()) < 5 and page.features.get("line_density", 0) > 0.02: | |
| flags.append("suspiciously_short") | |
| scores = sorted(b.confidence for b in result.blocks if b.confidence is not None) | |
| if scores and scores[int((len(scores) - 1) * 0.1)] < threshold: | |
| flags.append("low_confidence") | |
| lines = [s.strip() for s in result.text.splitlines() if s.strip()] | |
| if len(lines) > 5 and max(Counter(lines).values()) > len(lines) * 0.6: | |
| flags.append("repeated_text") | |
| if result.text.count("```") % 2: | |
| flags.append("markdown_unbalanced") | |
| if "\ufffd" in result.text: | |
| flags.append("unresolved_noise") | |
| if page.features.get("possible_columns") and result.source == "cpu_ocr": | |
| flags.append("reading_order_unverified") | |
| return list(dict.fromkeys(result.flags + flags)) | |