如何遍历Spacy中Meal标签Span并找出其完全包含的子Span？

首先，基于你提供的示例代码，我们可以通过以下步骤实现需求：获取所有Meal标签的Span，再逐一检查其他Span是否完全处于其边界范围内。

完整代码示例 #

import spacy
from spacy.tokens import Span

sent = 'I eat 5 apples and 2 bananas.'
nlp = spacy.load("en_core_web_sm")
doc = nlp(sent)

# 初始化自定义Span集合
doc.spans['sc'] = [
   Span(doc, 2, 3, 'Ingredient'),
   Span(doc, 5, 6, 'Ingredient'),
   Span(doc, 2, 6, 'Meal')]

# 1. 筛选所有标签为'Meal'的Span
meal_spans = [span for span in doc.spans['sc'] if span.label_ == 'Meal']

# 2. 遍历每个Meal Span，找出完全包含在其中的Span
for meal_span in meal_spans:
    print(f"当前Meal Span: {meal_span.text} (位置范围: {meal_span.start}-{meal_span.end})")
    # 筛选条件：排除自身 + 起始位置≥Meal Span起始 + 结束位置≤Meal Span结束
    contained_spans = [
        span for span in doc.spans['sc']
        if span != meal_span 
        and span.start >= meal_span.start 
        and span.end <= meal_span.end
    ]
    # 输出结果
    print("完全包含的子Span:")
    for span in contained_spans:
        print(f"- 内容: {span.text} | 标签: {span.label_} | 位置: {span.start}-{span.end}")

运行结果 #

当前Meal Span: 5 apples and 2 bananas (位置范围: 2-6)
完全包含的子Span:
- 内容: apples | 标签: Ingredient | 位置: 2-3
- 内容: bananas | 标签: Ingredient | 位置: 5-6

关键说明 #

• 核心判断逻辑通过span.start >= meal_span.start and span.end <= meal_span.end确保子Span完全处于Meal Span的边界内
• 用span != meal_span排除Meal Span自身，避免无效匹配
• 如果你的Span分散在doc.spans的多个key下，可以先合并所有Span再处理：all_spans = [span for spans in doc.spans.values() for span in spans]

内容的提问来源于stack exchange，提问作者user9092346