File size: 2,341 Bytes
188f4d8 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 | """字符提取（Parsing）任务。"""
from __future__ import annotations
import re
from ._text import clean_value, extract_by_prefix, strip_code_fence, strip_thinking
PROMPT = """你将被提供一张包含汉字字符的图片。请仔细观察图片，并将图片中的所有汉字字符以正确的阅读顺序提取出来。
要求：
- 严格按图片中文字的阅读顺序输出
- 仅输出识别到的字符本身，不要输出任何解释、标点补充或分析
输出格式（必须严格遵守）：
提取文本：<识别到的汉字字符>
"""
_OTHER_PREFIX_RE = re.compile(r"(?:字体分类|字体|分类)\s*[:：]")
_PREAMBLE_LINE_RE = re.compile(
r"^(?:好的|好|没问题|当然|根据(?:图片|图像)|这是|以下是|图片中(?:的)?(?:内容|文字|字符)?(?:为|是|如下)?|无法(?:识别|看清)|抱歉|对不起)"
)
def _fallback_full_answer(text: str) -> str:
"""模型不按格式输出时，把整段答案当作字符提取结果，但清理"开场白"与"其它任务前缀"。"""
if not text:
return ""
raw = strip_code_fence(text.strip())
if not raw:
return ""
cleaned: list[str] = []
for ln in raw.splitlines():
s = ln.strip().strip("`").strip()
if not s:
continue
if _OTHER_PREFIX_RE.match(s):
continue
if _PREAMBLE_LINE_RE.match(s):
continue
m = _OTHER_PREFIX_RE.search(s)
if m:
s = s[: m.start()].rstrip()
if not s:
continue
cleaned.append(s)
if not cleaned:
return ""
candidate = "\n".join(cleaned).strip()
candidate = clean_value(candidate) or candidate
if len(candidate) < 2:
return ""
return candidate
def extract(answer: str) -> tuple[bool, dict[str, str | bool]]:
text = strip_thinking(answer)
extracted = extract_by_prefix(
text,
["提取文本", "提取结果", "识别结果", "识别文本", "文本"],
merge_trailing_lines=True,
)
data: dict[str, str | bool] = {}
if extracted:
data["extracted_text"] = extracted
return True, data
fb = _fallback_full_answer(text)
if fb:
data["extracted_text"] = fb
data["fallback"] = True
return True, data
return False, data
|