docvortex 0.4.5__tar.gz → 0.4.6__tar.gz

{docvortex-0.4.5/src/docvortex.egg-info → docvortex-0.4.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: docvortex
-Version: 0.4.5
+Version: 0.4.6
 Summary: A fast, multi-format document parsing and conversion engine
 License-Expression: MIT
 Project-URL: Repository, https://github.com/myhloli/DocVortex
@@ -54,7 +54,7 @@ Dynamic: license-file
 
 <div align="center">
 
-<img src="https://raw.githubusercontent.com/myhloli/DocVortex/main/docs/images/docvortex-logo.jpg" alt="DocVortex logo" width="200">
+<img src="https://gcore.jsdelivr.net/gh/myhloli/DocVortex@main/docs/images/docvortex-logo.jpg" alt="DocVortex logo" width="200">
 
 # DocVortex
 
@@ -85,7 +85,7 @@ then exports the result in the formats your workflow needs.
 Native parsing works without an OCR or VLM inference service. Use DocVortex
 directly through its CLI or Python SDK, independently of MinerU.
 
-![DocVortex pipeline: native documents become a unified representation, then Markdown, HTML, LaTeX, DOCX, EPUB, PDF or structured JSON.](https://raw.githubusercontent.com/myhloli/DocVortex/main/docs/images/docvortex-overview.jpg)
+![DocVortex pipeline: native documents become a unified representation, then Markdown, HTML, LaTeX, DOCX, EPUB, PDF or structured JSON.](https://gcore.jsdelivr.net/gh/myhloli/DocVortex@main/docs/images/docvortex-overview.jpg)
 
 ## Quick start
 

{docvortex-0.4.5 → docvortex-0.4.6}/README.md

@@ -1,6 +1,6 @@
 <div align="center">
 
-<img src="https://raw.githubusercontent.com/myhloli/DocVortex/main/docs/images/docvortex-logo.jpg" alt="DocVortex logo" width="200">
+<img src="https://gcore.jsdelivr.net/gh/myhloli/DocVortex@main/docs/images/docvortex-logo.jpg" alt="DocVortex logo" width="200">
 
 # DocVortex
 
@@ -31,7 +31,7 @@ then exports the result in the formats your workflow needs.
 Native parsing works without an OCR or VLM inference service. Use DocVortex
 directly through its CLI or Python SDK, independently of MinerU.
 
-![DocVortex pipeline: native documents become a unified representation, then Markdown, HTML, LaTeX, DOCX, EPUB, PDF or structured JSON.](https://raw.githubusercontent.com/myhloli/DocVortex/main/docs/images/docvortex-overview.jpg)
+![DocVortex pipeline: native documents become a unified representation, then Markdown, HTML, LaTeX, DOCX, EPUB, PDF or structured JSON.](https://gcore.jsdelivr.net/gh/myhloli/DocVortex@main/docs/images/docvortex-overview.jpg)
 
 ## Quick start
 

{docvortex-0.4.5 → docvortex-0.4.6}/src/docvortex/analyzers/native/pdf/_script_geometry.py

@@ -620,6 +620,61 @@ def _assign_component(
         roles[index] = "body"
 
 
+def paired_script_roles(
+    chars: list[Char],
+    roles: list[ScriptRole],
+    tight_bboxes: dict[int, BBox],
+    origins: dict[int, tuple[float, float]],
+) -> dict[int, ScriptRole]:
+    """确认同一基字符右侧横向叠放的上下标，排除分式和远距邻列。"""
+
+    features = build_script_features(chars, tight_bboxes, origins, set())
+    visible = [feature for feature in features if feature.text.isprintable() and not feature.text.isspace()]
+    paired: dict[int, ScriptRole] = {}
+    for position, (base, first, second) in enumerate(zip(visible, visible[1:], visible[2:])):
+        if (
+            not all(feature.is_valid and feature.text.isalnum() for feature in (base, first, second))
+            or roles[base.index] != "body"
+            or {roles[first.index], roles[second.index]} != {"sup", "sub"}
+        ):
+            continue
+        assert base.tight_bbox is not None and base.origin is not None
+        assert first.tight_bbox is not None and first.origin is not None
+        assert second.tight_bbox is not None and second.origin is not None
+        scale = base.tight_height
+        if scale <= 0 or any(feature.tight_height > 0.8 * scale for feature in (first, second)):
+            continue
+        overlap = min(first.tight_bbox[2], second.tight_bbox[2]) - max(first.tight_bbox[0], second.tight_bbox[0])
+        minimum_width = min(first.tight_bbox[2] - first.tight_bbox[0], second.tight_bbox[2] - second.tight_bbox[0])
+        if overlap < 0.5 * minimum_width:
+            continue
+        if any(
+            not -0.15 * scale <= feature.tight_bbox[0] - base.tight_bbox[2] <= 0.5 * scale
+            or abs(feature.origin[1] - base.origin[1]) < 0.15 * scale
+            or abs(feature.origin[1] - base.origin[1]) > scale
+            for feature in (first, second)
+        ):
+            continue
+        if (first.origin[1] - base.origin[1]) * (second.origin[1] - base.origin[1]) >= 0:
+            continue
+        # 本补证据只接纳完整单字符角标，不能只给 grid 等多字母角标的首字母加样式。
+        following = visible[position + 3] if position + 3 < len(visible) else None
+        if (
+            following is not None
+            and following.is_valid
+            and following.text.isalnum()
+            and following.tight_bbox is not None
+            and following.origin is not None
+            and following.tight_height <= 0.8 * scale
+            and abs(following.origin[1] - second.origin[1]) <= 0.15 * scale
+            and -0.15 * scale <= following.tight_bbox[0] - second.tight_bbox[2] <= 0.5 * scale
+        ):
+            continue
+        paired[first.index] = roles[first.index]
+        paired[second.index] = roles[second.index]
+    return paired
+
+
 def classify_char_script_roles(
     chars: list[Char],
     *,

{docvortex-0.4.5 → docvortex-0.4.6}/src/docvortex/analyzers/native/pdf/_table_recovery/engine.py

@@ -19,6 +19,7 @@ from .contracts import (
 from .geometry import bbox_intersection, normalize_bbox
 from .sparse_hybrid import build_sparse_hybrid_candidates, diagnose_sparse_hybrid_candidate_builds
 from .sparse_multiline import build_sparse_multiline_candidates, diagnose_sparse_multiline_candidate_builds
+from .rule_band import build_rule_band_candidates
 from .text import build_native_table_text
 from .text_grid import build_text_candidates, diagnose_text_candidate_builds
 from .vector import MAX_PRIMITIVES_PER_TABLE, build_vector_candidates, diagnose_vector_candidate_builds
@@ -379,8 +380,11 @@ def _evaluate_native_pdf_table(
             )
         if _has_alias_affected_physical_blank_row(vector_attempts):
             sparse_hybrid_allowed = False
-    sparse_hybrid_candidates = build_sparse_hybrid_candidates(table_input, text) if sparse_hybrid_allowed else []
-    sparse_hybrid_selection = _select_candidate(sparse_hybrid_candidates)
+    rule_band_candidates = build_rule_band_candidates(table_input, text) if sparse_hybrid_allowed else []
+    sparse_hybrid_candidates = (
+        build_sparse_hybrid_candidates(table_input, text) if sparse_hybrid_allowed and not rule_band_candidates else []
+    )
+    sparse_hybrid_selection = _select_candidate([*rule_band_candidates, *sparse_hybrid_candidates])
     text_candidates = (
         build_text_candidates(table_input, text)
         if len(text.rows) >= 2 and vector_selection is None and sparse_hybrid_selection is None
@@ -396,6 +400,7 @@ def _evaluate_native_pdf_table(
             text_candidates = []
     existing_generated_candidates = [
         *vector_candidates,
+        *rule_band_candidates,
         *sparse_hybrid_candidates,
         *text_candidates,
     ]
@@ -466,7 +471,10 @@ def diagnose_native_pdf_table(table_input: NativeTableInput) -> dict[str, Any]:
         if evaluation.text is not None
         else ()
     )
-    sparse_multiline_attempts = (
+    rule_band_attempts: list[dict[str, Any]] = []
+    if evaluation.text is not None:
+        build_rule_band_candidates(table_input, evaluation.text, diagnostics=rule_band_attempts)
+    sparse_multiline_attempts = tuple(rule_band_attempts) + (
         diagnose_sparse_multiline_candidate_builds(
             table_input,
             evaluation.text,

docvortex-0.4.6/src/docvortex/analyzers/native/pdf/_table_recovery/rule_band.py

@@ -0,0 +1,135 @@
+"""以三线表物理表头边界恢复换行表头和同步重启的描述列。"""
+
+from __future__ import annotations
+
+from typing import Any
+
+from .candidate import GridCellSpec, build_candidate
+from .contracts import NativeTableCandidate, NativeTableInput, NativeTableText
+from .geometry import normalize_angle, table_local_size
+from .sparse_common import _local_rules
+from .sparse_multiline import (
+    _build_column_hypothesis,
+    _has_overlapping_formula_rows,
+    _local_rectangles,
+    _row_occupancy,
+    _stable_gutters,
+)
+
+
+def build_rule_band_candidates(
+    table_input: NativeTableInput,
+    text: NativeTableText,
+    diagnostics: list[dict[str, Any]] | None = None,
+) -> list[NativeTableCandidate]:
+    """只在三条完整横线及无歧义列占用共同成立时恢复特殊三线表。"""
+
+    width, height = table_local_size(table_input.table_bbox, normalize_angle(table_input.angle))
+    rules = _local_rules(table_input, width, height)
+    tolerance = max(1.0, 0.15 * text.median_glyph_height)
+    horizontal = sorted(
+        {rule.coordinate for rule in rules if rule.orientation == "horizontal" and rule.end - rule.start >= 0.9 * width}
+    )
+    if (
+        len(horizontal) != 3
+        or horizontal[0] > tolerance
+        or height - horizontal[-1] > tolerance
+        or any(rule.orientation == "vertical" and tolerance < rule.coordinate < width - tolerance for rule in rules)
+    ):
+        return []
+    header_bottom = horizontal[1]
+    header = [row for row in text.rows if row.bbox[3] <= header_bottom + 0.5]
+    body = [row for row in text.rows if row.bbox[1] >= header_bottom - 0.5]
+    if not 1 <= len(header) <= 3 or len(body) < 3 or len(header) + len(body) != len(text.rows):
+        return []
+    if _has_overlapping_formula_rows(text, header_bottom):
+        return []
+    hypothesis = _build_column_hypothesis(text, width, rules, _local_rectangles(table_input, width, height), None)
+    if hypothesis is None:
+        return []
+    tracks = hypothesis.x_tracks
+    cols = len(tracks) - 1
+    glyphs = {glyph.glyph_id: glyph for glyph in text.glyphs}
+    occupancy = [_row_occupancy(row, glyphs, tracks) for row in body]
+    full = set(range(cols))
+    common = set.intersection(*occupancy)
+    prefix_cols = min(common) if common else 0
+    groups: list[tuple[int, int]] = []
+    if prefix_cols:
+        # 至少两个描述列同步重启，右侧逐行完整，才能区分 rowspan 与普通空格。
+        suffix = set(range(prefix_cols, cols))
+        starts = [index for index, occupied in enumerate(occupancy) if occupied == full]
+        ends = starts[1:] + [len(body)]
+        if (
+            prefix_cols < 2
+            or len(suffix) < 2
+            or common != suffix
+            or len(starts) < 3
+            or starts[0] != 0
+            or any(occupied not in (full, suffix) for occupied in occupancy)
+            or any(end - start < 2 for start, end in zip(starts, ends))
+        ):
+            return []
+        groups = list(zip(starts, ends))
+    elif any(occupied != full for occupied in occupancy):
+        return []
+    if not groups and len(header) == 1:
+        # 普通单行表头三线表仍走既有策略，避免无关候选和金标漂移。
+        return []
+    header_occupancy = [_row_occupancy(row, glyphs, tracks) for row in header]
+    if set.union(*header_occupancy) != full:
+        return []
+    if len(header) > 1 and header_occupancy[-1] != full:
+        # 多级表头可能有仅出现在上层的 rowspan 标签，不能误折叠成普通换行表头。
+        return []
+    margin = max(0.25, 0.05 * text.median_glyph_width)
+    if any(
+        token.bbox[0] + margin < boundary < token.bbox[2] - margin
+        for row in header
+        for token in row.tokens
+        for boundary in tracks[1:-1]
+    ):
+        return []
+    if not _stable_gutters(text, tracks, hypothesis.physical_boundaries, header_bottom, None):
+        return []
+    boundaries = [header_bottom]
+    for previous, current in zip(body, body[1:]):
+        if previous.bbox[3] > current.bbox[1] + 0.5:
+            return []
+        boundaries.append((previous.bbox[3] + current.bbox[1]) / 2.0)
+    boundaries.append(height)
+    specs = [GridCellSpec(0, col, 1, 1, (tracks[col], 0.0, tracks[col + 1], header_bottom)) for col in range(cols)]
+    spans = {start: end for start, end in groups}
+    for index in range(len(body)):
+        for col in range(cols):
+            if col < prefix_cols and index not in spans:
+                continue
+            end = spans[index] if col < prefix_cols else index + 1
+            specs.append(
+                GridCellSpec(index + 1, col, end - index, 1, (tracks[col], boundaries[index], tracks[col + 1], boundaries[end]))
+            )
+    record: dict[str, Any] = {}
+    candidate = build_candidate(
+        source="sparse_multiline",
+        rows=len(body) + 1,
+        cols=cols,
+        specs=tuple(specs),
+        text=text,
+        structure_support=1.0,
+        row_stability=1.0,
+        column_stability=1.0,
+        issues=("evidence=physical_header_band", f"wrapped_header_rows={len(header)}", f"descriptor_columns={prefix_cols}"),
+        use_grid_index=True,
+        diagnostics=record,
+    )
+    if diagnostics is not None:
+        diagnostics.append({"source": "sparse_multiline", "evidence": "physical_header_band", **record})
+    if (
+        candidate is None
+        or candidate.text_capture < 1.0
+        or candidate.order_consistency < 1.0
+        or candidate.score < 0.98
+        or record.get("ambiguous_glyph_ratio", 1.0) > 0.0
+    ):
+        return []
+    return [candidate]

{docvortex-0.4.5 → docvortex-0.4.6}/src/docvortex/analyzers/native/pdf/_table_recovery/text.py

@@ -3,6 +3,7 @@
 from __future__ import annotations
 
 import statistics
+import re
 import unicodedata
 from dataclasses import dataclass
 
@@ -212,6 +213,9 @@ def _looks_like_compact_unit_or_identifier(
     combined = previous_line.strip() + next_line.strip()
     if not combined or any(char.isspace() for char in combined):
         return False
+    # 带单位后缀的长英文标题仍有词界，不能仅因末尾 /% 就把相邻单词粘连。
+    if re.fullmatch(r"[A-Za-z]{4,}", previous_line.strip()) and re.match(r"[A-Za-z]{4,}(?:/|%)", next_line.strip()):
+        return False
     return any(char.isdigit() or char in "/_^%°µμ" for char in combined)
 
 

{docvortex-0.4.5 → docvortex-0.4.6}/src/docvortex/analyzers/native/pdf/auxiliary_text.py

@@ -336,11 +336,22 @@ def _classify_page_footnotes(
             # 图形坐标轴和外框不能充当页面脚注分隔线。
             continue
         rule_source_indices: set[int] = set()
-        for lane in lanes:
+        rule_lanes = list(lanes)
+        rule_width = axis_line.bbox[2] - axis_line.bbox[0]
+        # 通栏长线自身提供局部栏界；上方双栏布局不能拆散线下的整页脚注。
+        if rule_width >= 0.7 * local_page_width:
+            corridor = [
+                (line, bbox)
+                for line, bbox in line_geometry
+                if bbox[0] >= axis_line.bbox[0] - 0.25 * median_height and bbox[2] <= axis_line.bbox[2] + 0.25 * median_height
+            ]
+            rule_lanes.append(_TextLane(axis_line.bbox[0], axis_line.bbox[2], corridor))
+        for lane in rule_lanes:
             following_rule_tops = [
                 other.bbox[1]
                 for other in local_axis_lines
                 if other.orientation == "horizontal"
+                and other.bbox[2] - other.bbox[0] >= max(4.0 * median_height, 0.04 * local_page_width)
                 and other.bbox[1] - axis_line.bbox[3] > 0.5 * median_height
                 and _bbox_axis_overlap_ratio(
                     axis_line.bbox,
@@ -1652,6 +1663,9 @@ def _chinese_page_number_to_int(value: str) -> int | None:
 def _marginal_text_matches(first_text: str, second_text: str) -> bool:
     """在屏蔽变化数字后比较边缘稳定文本，短文本只接受完全一致。"""
 
+    # 公式编号在数字屏蔽后都会成为同一标记，不能作为重复页脚的文本证据。
+    if any(re.fullmatch(r"[（(﹙]\s*[A-Za-z]?\d+(?:[.\-]\d+)*\s*[)）﹚]", text.strip()) for text in (first_text, second_text)):
+        return False
     first = _normalize_marginal_text(first_text)
     second = _normalize_marginal_text(second_text)
     if not first or not second:

{docvortex-0.4.5 → docvortex-0.4.6}/src/docvortex/analyzers/native/pdf/formulas.py

@@ -9,6 +9,7 @@ from dataclasses import dataclass, replace
 from typing import Any
 
 from ....document.pdf._document import PDFPathInfo
+from ....document.pdf.text._contracts import Char
 from ....foundation._text import build_tagged_formula_content
 from ....schema import BBox
 from .geometry import (
@@ -715,6 +716,13 @@ def _build_formula_like_blocks(
                 blocks.append(block)
                 claimed_source_indices.update(line.source_index for line, _bbox in members)
 
+    recovered, recovered_indices = _build_mixed_body_display_formulas(
+        [line for line in lines if line.source_index not in claimed_source_indices],
+        table_bboxes,
+        page_size,
+    )
+    blocks.extend(recovered)
+    claimed_source_indices.update(recovered_indices)
     remaining_lines = [
         line
         for line in lines
@@ -724,6 +732,121 @@ def _build_formula_like_blocks(
     return blocks, remaining_lines + paragraph_lines
 
 
+def _build_mixed_body_display_formulas(
+    lines: list[_LineItem],
+    table_bboxes: list[BBox],
+    page_size: tuple[float, float],
+) -> tuple[list[dict[str, Any]], set[int]]:
+    """在混合字体正文的独立留白带内恢复无编号公式，避免把碎片窄栏当作正文栏。"""
+
+    blocks: list[dict[str, Any]] = []
+    claimed: set[int] = set()
+    for angle in sorted({line.angle for line in lines}):
+        geometry = [
+            (line, _rotate_bbox_to_upright(line.bbox, page_size, angle))
+            for line in lines
+            if line.angle == angle and line.semantic_type is None and not line.paragraph_formula_context
+        ]
+        width = page_size[1] if angle in {90, 270} else page_size[0]
+        body = [
+            (line, bbox) for line, bbox in geometry if bbox[2] - bbox[0] >= 0.5 * width and _formula_prefix_has_prose(line.text)
+        ]
+        # 该补充路径只处理既有 dominant-font 路径缺少稳定正文覆盖的混排页面。
+        if sum(line.font_coverage < 0.75 for line, _bbox in body) < 2:
+            continue
+        height = statistics.median(_line_effective_height(line, bbox) for line, bbox in body)
+        body_left = statistics.median(bbox[0] for _line, bbox in body)
+        body_right = statistics.median(bbox[2] for _line, bbox in body)
+        body_width = body_right - body_left
+        barriers = sorted(
+            [
+                (line, bbox)
+                for line, bbox in geometry
+                if _formula_prefix_has_prose(line.text)
+                and (bbox[2] - bbox[0] >= 0.5 * body_width or len(re.findall(r"[\u3400-\u9fff]", line.text)) >= 2)
+            ],
+            key=lambda item: item[1][1],
+        )
+        for previous, following in zip(barriers, barriers[1:]):
+            top, bottom = previous[1][3], following[1][1]
+            if not 0.6 * height <= bottom - top <= 8.0 * height:
+                continue
+            members = [
+                (line, bbox)
+                for line, bbox in geometry
+                if line.source_index not in {previous[0].source_index, following[0].source_index}
+                and line.source_index not in claimed
+                and bbox[1] >= top - 0.2 * height
+                and bbox[3] <= bottom + 0.2 * height
+                and bbox[0] >= body_left
+                and bbox[2] <= body_right
+            ]
+            if not members:
+                continue
+            bbox = _bbox_union_many([box for _line, box in members])
+            text = " ".join(line.text for line, _box in members)
+            if (
+                not any(char in "=∑∫√≤≥<>" for char in text)
+                or len(re.findall(r"[\u3400-\u9fff]", text)) >= 2
+                or len(re.findall(r"\b[A-Za-z]{4,}\b", text)) >= 2
+                or not 0.15 * body_width <= bbox[2] - bbox[0] <= 0.8 * body_width
+                or abs(_bbox_center_x(bbox) - 0.5 * (body_left + body_right)) > 0.1 * body_width
+                or any(
+                    _bbox_overlap_in_smaller(line.bbox, table_bbox) > 0 for line, _box in members for table_bbox in table_bboxes
+                )
+            ):
+                continue
+            block = _formula_members_to_block(members, page_size, angle, anchor_source_index=members[0][0].source_index)
+            if block is not None:
+                blocks.append(block)
+                claimed.update(line.source_index for line, _bbox in members)
+    return blocks, claimed
+
+
+def _unmapped_formula_ink_bboxes(chars: list[Char]) -> list[BBox]:
+    """保留实际绘制、但映射为空白或控制码的高字形几何，不改变公开文本。"""
+
+    output: list[BBox] = []
+    for char in chars:
+        value = str(char.get("char", ""))
+        if value.isprintable() and not value.isspace():
+            continue
+        bbox = _coerce_bbox(char.get("tight_bbox"))
+        font = char.get("font") or {}
+        size = float(font.get("size", 0) or 0)
+        if (
+            bbox is not None
+            and size > 0
+            and char.get("text_object_id") is not None
+            and char.get("text_render_mode") not in {3, 7}
+            and bbox[2] - bbox[0] > 0.05 * size
+            and bbox[3] - bbox[1] > 1.5 * size
+        ):
+            output.append(bbox)
+    return output
+
+
+def _attach_unmapped_formula_ink(blocks: list[dict[str, Any]], bboxes: list[BBox]) -> None:
+    """将紧贴公式且纵向相容的未映射字形唯一认领到公式框。"""
+
+    for bbox in bboxes:
+        matches = [
+            block
+            for block in blocks
+            if block.get("type") == "equation"
+            and int(block.get("angle", 0)) in {0, 180}
+            and _bbox_axis_overlap_ratio(bbox, block["bbox"], axis="y") >= 0.8
+            and _bbox_distance(bbox, block["bbox"]) <= 0.25 * (bbox[3] - bbox[1])
+            and bbox[1] >= block["bbox"][1] - 2.0
+            and bbox[3] <= block["bbox"][3] + 2.0
+        ]
+        if len(matches) == 1:
+            block = matches[0]
+            block["bbox"] = _bbox_union(block["bbox"], bbox)
+            if block.get("_tight_output_bbox") is not None:
+                block["_tight_output_bbox"] = _bbox_union(block["_tight_output_bbox"], bbox)
+
+
 def _formula_line_has_math_operator(text: str) -> bool:
     """检查文本行是否具有独立公式常见的数学运算符。"""
 

{docvortex-0.4.5 → docvortex-0.4.6}/src/docvortex/analyzers/native/pdf/graphics.py

@@ -60,6 +60,21 @@ def _form_supersedes_nested_bbox(form_bbox: BBox, nested_bbox: BBox) -> bool:
     return form_area > 0 and nested_area < 0.5 * form_area and _bbox_overlap_in_first(nested_bbox, form_bbox) >= 0.9
 
 
+def _form_member_bbox(line: _LineItem, form_bbox: BBox) -> BBox | None:
+    """优先保持既有行框，仅在 Form 边缘用完整的可见字形证据补回成员。"""
+
+    if _bbox_overlap_in_first(line.bbox, form_bbox) >= 0.9:
+        return line.bbox
+    ink_bbox = line.ink_bbox
+    if ink_bbox is None:
+        visible_chars = [char for char in line.chars if str(char.get("char", "")).strip()]
+        tight = [_coerce_bbox(char.get("tight_bbox")) for char in visible_chars]
+        if not tight or any(bbox is None for bbox in tight):
+            return None
+        ink_bbox = _bbox_union_many([bbox for bbox in tight if bbox is not None])
+    return ink_bbox if _bbox_overlap_in_first(ink_bbox, form_bbox) >= 0.99 else None
+
+
 def _tighten_form_image_bbox(
     source: _PageSource,
     form_bbox: BBox,
@@ -79,7 +94,7 @@ def _tighten_form_image_bbox(
     # 至少两个嵌套 Path 和四个矢量元素，避免只凭普通边框或少量文本裁剪 Form。
     if len(internal_paths) < 2 or len(internal_paths) + len(internal_drawing_lines) < 4:
         return form_bbox
-    internal_text = [line.bbox for line in source.lines if _bbox_overlap_in_first(line.bbox, form_bbox) >= 0.9]
+    internal_text = [bbox for line in source.lines if (bbox := _form_member_bbox(line, form_bbox)) is not None]
     evidence_bbox = _clip_bbox(
         _bbox_union_many(internal_paths + internal_drawing_lines + internal_text),
         source.page_size,
@@ -125,7 +140,7 @@ def _select_form_image_bboxes(source: _PageSource) -> list[BBox]:
         member_rows = {
             line.visual_row_id if line.visual_row_id is not None else line.source_index
             for line in source.lines
-            if _bbox_overlap_in_first(line.bbox, bbox) >= 0.9
+            if _form_member_bbox(line, bbox) is not None
         }
         internal_drawing_count = sum(
             _bbox_overlap_in_first(drawing_line.bbox, bbox) >= 0.9 for drawing_line in source.drawing_lines
@@ -151,9 +166,7 @@ def _build_form_image_blocks(
         if line.source_index in claimed_line_indices:
             continue
         matching_indices = [
-            candidate_index
-            for candidate_index, bbox in enumerate(form_bboxes)
-            if _bbox_overlap_in_first(line.bbox, bbox) >= 0.9
+            candidate_index for candidate_index, bbox in enumerate(form_bboxes) if _form_member_bbox(line, bbox) is not None
         ]
         if not matching_indices:
             continue

{docvortex-0.4.5 → docvortex-0.4.6}/src/docvortex/analyzers/native/pdf/inline/detection.py

@@ -225,6 +225,13 @@ def _drawing_match_for_line(
     hit_chars = [char for char in line.visible_chars if drawing_bbox[0] <= (char.bbox[0] + char.bbox[2]) / 2 <= drawing_bbox[2]]
     if not hit_chars:
         return None
+    if style == "strikethrough":
+        # 分式横线可能穿过字体外框却完全位于字形上方，不应据此生成删除线。
+        ink_boxes = [_coerce_bbox(line.chars[char.source_index].get("tight_bbox")) for char in hit_chars]
+        if all(bbox is not None for bbox in ink_boxes) and not any(
+            bbox[1] <= drawing_center_y <= bbox[3] for bbox in ink_boxes if bbox is not None
+        ):
+            return None
     hit_left = min(char.bbox[0] for char in hit_chars)
     hit_right = max(char.bbox[2] for char in hit_chars)
     if (hit_right - hit_left) / drawing_length < TEXT_DECORATION_MIN_TEXT_COVERAGE_RATIO:

{docvortex-0.4.5 → docvortex-0.4.6}/src/docvortex/analyzers/native/pdf/inline/scripts.py

@@ -7,7 +7,7 @@ import unicodedata
 from typing import Any, Literal, Sequence
 
 from .....schema import BBox
-from .._script_geometry import ScriptRole, classify_char_script_roles
+from .._script_geometry import ScriptRole, classify_char_script_roles, paired_script_roles
 from ..geometry import _rotate_bbox_to_upright
 from .common import _coerce_bbox, _normalize_match_fragment, _ordered_line_chars
 from .types import (
@@ -678,7 +678,7 @@ def _strong_structural_script_roles(
         tight_bboxes=tight_bboxes,
         origins=origins,
     )
-    strong_roles: dict[int, ScriptRole] = {}
+    strong_roles: dict[int, ScriptRole] = paired_script_roles(chars, roles, tight_bboxes, origins)
     for index in _citation_script_indices(chars, roles):
         if roles[index] != "body":
             strong_roles[index] = roles[index]

{docvortex-0.4.5 → docvortex-0.4.6}/src/docvortex/analyzers/native/pdf/line_merging.py

@@ -18,6 +18,7 @@ from .geometry import (
 from .line_layout import _connection_crosses_table, _font_signatures_share_family, _infer_text_lanes, _line_effective_height
 from .models import _LineItem, _TextLane
 from .native_text import _fill_native_typography, _median_native_glyph_width
+from ._script_geometry import classify_char_script_roles, paired_script_roles
 
 
 def _merge_same_baseline_text_lines(
@@ -273,13 +274,35 @@ def _classify_overlapping_inline_cluster(
         and line.font_coverage >= 0.75
         for line, bbox in members
     )
-    if has_body_host:
+    if has_body_host or _has_paired_script_body_host(members, median_height):
         return "inline"
     if len(members) >= 3 and len(visual_row_ids) >= 3 and union_bbox[2] - union_bbox[0] <= 0.6 * lane_width:
         return "formula"
     return None
 
 
+def _has_paired_script_body_host(members: list[tuple[_LineItem, BBox]], median_height: float) -> bool:
+    """用正文宿主和跨 run 的成对上下标补证据，不放宽普通同行重叠阈值。"""
+
+    if len(members) != 2 or any(line.angle != 0 or line.preserve_split_boundary or line.semantic_type for line, _ in members):
+        return False
+    host, bbox = max(members, key=lambda item: item[1][2] - item[1][0])
+    has_prose = len(re.findall(r"[\u3400-\u9fff]", host.text)) >= 2 or len(re.findall(r"\b[A-Za-z]{2,}\b", host.text)) >= 2
+    if not has_prose or bbox[2] - bbox[0] < 4.0 * median_height:
+        return False
+    owned_chars = [(char, line.source_index) for line, _ in members for char in line.chars]
+    if not owned_chars or any(not isinstance(char.get("char_idx"), int) for char, _ in owned_chars):
+        return False
+    owned_chars.sort(key=lambda item: item[0]["char_idx"])
+    chars = [char for char, _ in owned_chars]
+    tight = {char["char_idx"]: char["tight_bbox"] for char in chars if char.get("tight_bbox") is not None}
+    origins = {char["char_idx"]: char["origin"] for char in chars if char.get("origin") is not None}
+    roles = classify_char_script_roles(chars, tight_bboxes=tight, origins=origins)
+    paired = paired_script_roles(chars, roles, tight, origins)
+    # 上下标必须分属两个待合并 run，避免把一侧已有角标当作吞并邻块的依据。
+    return len(paired) == 2 and len({owned_chars[index][1] for index in paired}) == 2
+
+
 def _select_overlapping_inline_cluster_host(
     members: list[tuple[_LineItem, BBox]],
     median_height: float,
@@ -335,8 +358,10 @@ def _merge_overlapping_inline_cluster(
         if compact_formula_cluster
         else [line.bbox for line in ordered_members if line is not host]
     )
+    # 成对角标跨 run 只是 PDF 物理拆行，不应在上、下标之间补正文空格。
+    separator = "" if _has_paired_script_body_host(members, median_height) else " "
     merged = _LineItem(
-        text=" ".join(text for line in ordered_members if (text := line.text.strip())),
+        text=separator.join(text for line in ordered_members if (text := line.text.strip())),
         bbox=_bbox_union_many([line.bbox for line in ordered_members]),
         angle=host.angle,
         source_index=min(line.source_index for line in ordered_members),

{docvortex-0.4.5 → docvortex-0.4.6}/src/docvortex/analyzers/native/pdf/models.py

@@ -189,6 +189,7 @@ class _PreparedPage:
     page_footnote_groups: list[set[int]] = field(default_factory=list)
     script_lines: list[PDFTextScriptLine] = field(default_factory=list)
     formula_candidate_lines: list[_LineItem] = field(default_factory=list)
+    formula_ink_bboxes: list[BBox] = field(default_factory=list)
 
 
 @dataclass(slots=True)

{docvortex-0.4.5 → docvortex-0.4.6}/src/docvortex/analyzers/native/pdf/pipeline.py

@@ -30,7 +30,12 @@ from .auxiliary_text import (
 )
 from .char_geometry import DocumentGeometryPlan, apply_line_geometry_repairs, build_document_geometry_plan
 from .code_blocks import _build_code_blocks, _build_rule_delimited_code_blocks
-from .formulas import _build_formula_like_blocks, _build_vector_formula_blocks
+from .formulas import (
+    _attach_unmapped_formula_ink,
+    _build_formula_like_blocks,
+    _build_vector_formula_blocks,
+    _unmapped_formula_ink_bboxes,
+)
 from .geometry import (
     _bbox_area,
     _bbox_axis_overlap_ratio,
@@ -706,6 +711,7 @@ def _prepare_page_source(
         canonical_formula_source_lines=canonical_formula_source_lines,
         script_lines=script_lines,
         formula_candidate_lines=formula_candidate_lines,
+        formula_ink_bboxes=_unmapped_formula_ink_bboxes(source.chars),
     )
     _classify_page_auxiliary_text(prepared)
     return prepared
@@ -880,6 +886,7 @@ def _finalize_prepared_page(
         ) == _formula_block_inventory(formula_blocks):
             formula_blocks = canonical_formula_blocks
     prepared.canonical_formula_source_lines.clear()
+    _attach_unmapped_formula_ink(formula_blocks, prepared.formula_ink_bboxes)
     if prepared.canonical_formula_geometry:
         for line in remaining_lines:
             line.style_scale_repaired = original_style_scale_state.get(