📖 第23课：事件抽取

import re from collections import defaultdict class EventExtractor: def __init__(self): self.event_patterns = { '人事变动': [ r'(.+?)(加入|入职|加盟)了?(.+)', r'(.+?)(离开|离职|退出)了?(.+)', r'(.+?)(升任|就任|担任)(.+)', ], '收购并购': [ r'(.+?)(收购|并购|兼并)了?(.+)', r'(.+?)(被)(.+?)(收购)', ], '产品发布': [ r'(.+?)(发布|推出|上市)了?(.+)', r'(.+?)(发布|推出)(.+)', ], '融资': [ r'(.+?)(获得|完成|拿到)(.+?)(融资|投资|融资轮)', r'(.+?)(融资|投资)(.+?)(美元|人民币|元)', ], } def extract(self, text): events = [] for event_type, patterns in self.event_patterns.items(): for pattern in patterns: match = re.search(pattern, text) if match: events.append({ 'type': event_type, 'args': [g.strip() for g in match.groups() if g], 'text': text }) break return events extractor = EventExtractor() test_texts = [ "张三加入了字节跳动", "阿里巴巴收购了饿了么", "苹果发布了iPhone15", "OpenAI推出了GPT-5", "某公司完成了A轮融资", "李四就任CTO", "王五离开腾讯", ] print("=== 事件抽取 ===") for text in test_texts: events = extractor.extract(text) if events: for event in events: args_str = ', '.join(event['args']) print(f" [{event['type']}] {args_str} ← {text}") else: print(f" 未识别: {text}")

class EventElementExtractor: def __init__(self): self.roles = { '人事变动': {'触发词': ['加入', '离职', '就任', '离开', '升任'], '角色': ['人物', '组织']}, '收购': {'触发词': ['收购', '并购', '兼并'], '角色': ['收购方', '被收购方']}, '产品发布': {'触发词': ['发布', '推出', '上市'], '角色': ['发布方', '产品']}, } def identify_event_type(self, text): for event_type, config in self.roles.items(): for trigger in config['触发词']: if trigger in text: return event_type, trigger return None, None def extract_elements(self, text, event_type, trigger): import jieba.posseg as pseg words = list(pseg.cut(text)) elements = {'触发词': trigger} if event_type == '人事变动': for w in words: if w.flag == 'nr': elements['人物'] = w.word if w.flag == 'nt': elements['组织'] = w.word elif event_type == '收购': parts = text.split(trigger) if len(parts) >= 2: elements['收购方'] = parts[0].strip() elements['被收购方'] = parts[1].replace('了', '').strip() elif event_type == '产品发布': parts = text.split(trigger) if len(parts) >= 2: elements['发布方'] = parts[0].strip() elements['产品'] = parts[1].replace('了', '').strip() return elements extractor = EventElementExtractor() test_texts = [ "张三加入了字节跳动", "阿里巴巴收购了饿了么", "苹果发布了iPhone15", ] print("=== 事件要素抽取 ===") for text in test_texts: event_type, trigger = extractor.identify_event_type(text) if event_type: elements = extractor.extract_elements(text, event_type, trigger) print(f" 文本: {text}") print(f" 事件类型: {event_type}") for role, value in elements.items(): print(f" {role}: {value}") print()