| |
| |
|
|
| from __future__ import annotations |
|
|
|
|
| import re |
|
|
|
|
| _RE_EXTRACT_CLASS_BBOX = re.compile( |
| r"<x_(\d+(?:\.\d+)?)><y_(\d+(?:\.\d+)?)>(.*?)<x_(\d+(?:\.\d+)?)><y_(\d+(?:\.\d+)?)><class_([^>]+)>", |
| re.DOTALL, |
| ) |
|
|
|
|
| def extract_classes_bboxes( |
| text: str, |
| ) -> tuple[list[str], list[tuple[float, float, float, float]], list[str]]: |
| classes: list[str] = [] |
| bboxes: list[tuple[float, float, float, float]] = [] |
| texts: list[str] = [] |
| for match in _RE_EXTRACT_CLASS_BBOX.finditer(text or ""): |
| x1, y1, body, x2, y2, cls = match.groups() |
| classes.append(cls) |
| bboxes.append((float(x1), float(y1), float(x2), float(y2))) |
| texts.append(body) |
| return classes, bboxes, texts |
|
|
|
|
| def convert_latex_tables_in_markdown(markdown_text: str) -> str: |
| if "\\begin{tabular}" not in markdown_text: |
| return markdown_text |
| try: |
| from .convert.latex2html import latex_table_to_html |
|
|
| return latex_table_to_html(markdown_text) |
| except (ValueError, IndexError): |
| return markdown_text |
|
|
|
|
| def mmd_to_markdown(prediction: str, text_transform=None) -> str: |
| classes, _boxes, texts = extract_classes_bboxes(prediction or "") |
| if not texts: |
| cleaned = re.sub(r"<x_[^>]+>|<y_[^>]+>|<class_[^>]+>", "", prediction or "") |
| return convert_latex_tables_in_markdown(cleaned.replace("<tbc>", "").strip()) |
|
|
| blocks = [] |
| for cls, text in zip(classes, texts): |
| if text_transform is not None: |
| text = text_transform(cls, text) |
| text = text.replace("<tbc>", "").strip() |
| if not text: |
| continue |
| if cls in {"Table", "Chart"}: |
| text = convert_latex_tables_in_markdown(text) |
| elif cls == "Title" and not text.startswith("#"): |
| text = f"# {text}" |
| elif cls == "Section-header" and not text.startswith("#"): |
| text = f"## {text}" |
| blocks.append(text) |
| return "\n\n".join(blocks).strip() |
|
|
