50 lines
1.9 KiB
Python
50 lines
1.9 KiB
Python
# -*- coding: utf-8 -*-
|
||
"""
|
||
基于规则的新闻数据提取（无需 Ollama）
|
||
从新闻文本中提取战损、报复情绪等数值，供 db_merge 写入
|
||
"""
|
||
import re
|
||
from datetime import datetime, timezone
|
||
from typing import Any, Dict, Optional
|
||
|
||
|
||
def _first_int(text: str, pattern: str) -> Optional[int]:
|
||
m = re.search(pattern, text, re.I)
|
||
if m and m.group(1) and m.group(1).replace(",", "").isdigit():
|
||
return max(0, int(m.group(1).replace(",", "")))
|
||
return None
|
||
|
||
|
||
def extract_from_news(text: str, timestamp: Optional[str] = None) -> Dict[str, Any]:
|
||
"""
|
||
规则提取：匹配数字+关键词，输出符合 panel schema 的字段（无需 Ollama）
|
||
"""
|
||
ts = timestamp or datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%S.000Z")
|
||
out: Dict[str, Any] = {}
|
||
t = (text or "").lower()
|
||
|
||
loss_us, loss_ir = {}, {}
|
||
v = _first_int(t, r"(?:us|american|u\.?s\.?)[\s\w]*(?:say|report)[\s\w]*(\d+)[\s\w]*(?:troop|soldier|killed|dead)")
|
||
if v is not None:
|
||
loss_us["personnel_killed"] = v
|
||
v = _first_int(t, r"(\d+)[\s\w]*(?:us|american)[\s\w]*(?:troop|soldier|killed|dead)")
|
||
if v is not None:
|
||
loss_us["personnel_killed"] = v
|
||
v = _first_int(t, r"(?:iran|iranian)[\s\w]*(?:say|report)[\s\w]*(\d+)[\s\w]*(?:troop|soldier|killed|dead)")
|
||
if v is not None:
|
||
loss_ir["personnel_killed"] = v
|
||
v = _first_int(t, r"(\d+)[\s\w]*(?:iranian|iran)[\s\w]*(?:troop|soldier|killed|dead)")
|
||
if v is not None:
|
||
loss_ir["personnel_killed"] = v
|
||
|
||
if loss_us:
|
||
out.setdefault("combat_losses_delta", {})["us"] = loss_us
|
||
if loss_ir:
|
||
out.setdefault("combat_losses_delta", {})["iran"] = loss_ir
|
||
if "retaliat" in t or "revenge" in t or "报复" in t:
|
||
out["retaliation"] = {"value": 75, "time": ts}
|
||
if "wall street" in t or " dow " in t or "s&p" in t or "market slump" in t or "stock fall" in t or "美股" in t:
|
||
out["wall_street"] = {"time": ts, "value": 55}
|
||
|
||
return out
|