🏗️ 第3课：本体建模(OWL)

class OWLClass: """OWL类""" def __init__(self, name, parent=None): self.name = name self.parent = parent # 父类 self.children = [] # 子类列表 self.equivalent = [] # 等价类 self.disjoint = [] # 不相交类 self.instances = [] # 实例列表 self.annotations = {} # 注释 if parent: parent.children.append(self) def add_instance(self, instance): self.instances.append(instance) def get_all_instances(self): """获取本类及所有子类的实例（继承推理）""" result = list(self.instances) for child in self.children: result.extend(child.get_all_instances()) return result def is_subclass_of(self, other): """判断是否是other的子类（传递推理）""" if self == other: return True if self.parent: return self.parent.is_subclass_of(other) return False def __repr__(self): return f"OWLClass('{self.name}')" class OWLProperty: """OWL属性""" def __init__(self, name, prop_type="object"): self.name = name self.prop_type = prop_type # "object" 或 "datatype" self.domain = [] # 定义域 self.range = [] # 值域 self.parent = None # 父属性 self.cardinality = {} # {class: (min, max)} self.inverse_of = None # 逆属性 self.functional = False # 函数属性（最多一个值） self.transitive = False # 传递属性 self.symmetric = False # 对称属性 def __repr__(self): return f"OWLProperty('{self.name}', {self.prop_type})" class OWLOntology: """OWL本体""" def __init__(self, name): self.name = name self.classes = {} # {name: OWLClass} self.properties = {} # {name: OWLProperty} self.individuals = {} # {name: {type, attrs}} self.axioms = [] # 公理列表 def add_class(self, name, parent_name=None): parent = self.classes.get(parent_name) if parent_name else None cls = OWLClass(name, parent) self.classes[name] = cls return cls def add_property(self, name, prop_type="object"): prop = OWLProperty(name, prop_type) self.properties[name] = prop return prop def add_individual(self, name, class_name, attrs=None): self.individuals[name] = { "type": class_name, "attrs": attrs or {} } if class_name in self.classes: self.classes[class_name].add_instance(name) return name def add_disjoint(self, class1, class2): """声明两个类不相交""" if class1 in self.classes and class2 in self.classes: self.classes[class1].disjoint.append(class2) self.classes[class2].disjoint.append(class1) self.axioms.append(("disjoint", class1, class2)) def check_consistency(self): """简单一致性检查""" issues = [] # 检查不相交类是否有共同实例 for name, cls in self.classes.items(): my_instances = set(cls.get_all_instances()) for disjoint_name in cls.disjoint: other_instances = set(self.classes[disjoint_name].get_all_instances()) overlap = my_instances & other_instances if overlap: issues.append(f"不一致: {overlap} 同时属于不相交类 '{name}' 和 '{disjoint_name}'") # 检查函数属性的基数约束 for name, prop in self.properties.items(): if prop.functional: # 统计每个主体对该属性的值数量 attr_counts = {} for ind_name, ind_data in self.individuals.items(): val = ind_data["attrs"].get(name) if val: attr_counts.setdefault(ind_name, set()).add(val) for ind, vals in attr_counts.items(): if len(vals) > 1: issues.append(f"基数违反: {ind} 对函数属性 '{name}' 有多个值: {vals}") return issues def infer_types(self, individual_name): """推理个体所属的所有类型（含父类推理）""" ind = self.individuals.get(individual_name) if not ind: return set() types = {ind["type"]} current = self.classes.get(ind["type"]) while current and current.parent: types.add(current.parent.name) current = current.parent return types def infer_inverse(self, subject, prop_name, obj): """逆属性推理""" prop = self.properties.get(prop_name) if prop and prop.inverse_of: inverse_name = prop.inverse_of return (obj, inverse_name, subject) return None def summary(self): """本体摘要""" return { "名称": self.name, "类数": len(self.classes), "属性数": len(self.properties), "实例数": len(self.individuals), "公理数": len(self.axioms) } # ========== 构建文学领域本体 ========== onto = OWLOntology("中国文学领域本体") # 类层次结构 onto.add_class("人物") onto.add_class("作家", "人物") onto.add_class("诗人", "作家") onto.add_class("小说家", "作家") onto.add_class("作品") onto.add_class("小说", "作品") onto.add_class("诗集", "作品") onto.add_class("地点") onto.add_class("城市", "地点") onto.add_class("省份", "地点") # 不相交声明 onto.add_disjoint("小说", "诗集") onto.add_disjoint("城市", "省份") # 对象属性 创作 = onto.add_property("创作") 创作.domain = [onto.classes["作家"]] 创作.range = [onto.classes["作品"]] 创作.inverse_of = "被创作" 被创作 = onto.add_property("被创作") 被创作.domain = [onto.classes["作品"]] 被创作.range = [onto.classes["作家"]] 被创作.inverse_of = "创作" 出生地 = onto.add_property("出生地") 出生地.domain = [onto.classes["人物"]] 出生地.range = [onto.classes["地点"]] 出生地.functional = True 位于 = onto.add_property("位于") 位于.domain = [onto.classes["城市"]] 位于.range = [onto.classes["省份"]] 位于.transitive = True 影响 = onto.add_property("影响") 影响.symmetric = True # 数据属性 原名 = onto.add_property("原名", "datatype") 原名.functional = True 出生年份 = onto.add_property("出生年份", "datatype") 出生年份.functional = True # 添加实例 onto.add_individual("鲁迅", "小说家", {"出生地": "绍兴", "原名": "周树人", "出生年份": "1881"}) onto.add_individual("老舍", "小说家", {"出生地": "北京", "原名": "舒庆春", "出生年份": "1899"}) onto.add_individual("徐志摩", "诗人", {"出生地": "海宁", "出生年份": "1897"}) onto.add_individual("呐喊", "小说", {"被创作": "鲁迅"}) onto.add_individual("再别康桥", "诗集", {"被创作": "徐志摩"}) onto.add_individual("绍兴", "城市", {}) onto.add_individual("北京", "城市", {}) onto.add_individual("浙江省", "省份", {}) # 输出本体摘要 print("=== 本体摘要 ===") for k, v in onto.summary().items(): print(f" {k}: {v}") # 类层次 print(" === 类层次结构 ===") def print_class_tree(cls, indent=0): print(" " * indent + f"├─ {cls.name} ({len(cls.instances)}实例)") for child in cls.children: print_class_tree(child, indent + 1) for cls in onto.classes.values(): if cls.parent is None: print_class_tree(cls) # 类型推理 print(" === 类型推理 ===") for name in ["鲁迅", "徐志摩"]: types = onto.infer_types(name) print(f" {name} 的所有类型: {types}") # 一致性检查 print(" === 一致性检查 ===") issues = onto.check_consistency() if issues: for issue in issues: print(f" ❌ {issue}") else: print(" ✅ 本体一致，无冲突") # 逆属性推理 print(" === 逆属性推理 ===") inv = onto.infer_inverse("鲁迅", "创作", "呐喊") if inv: print(f" (鲁迅, 创作, 呐喊) → 推理: ({inv[0]}, {inv[1]}, {inv[2]})")

# N-元关系模式实现 class NaryRelation: """N-元关系：用中间节点表示多元关系""" def __init__(self, ontology): self.onto = ontology self.relations = {} # {relation_id: {role: entity}} def create(self, relation_type, roles): """创建N-元关系 Args: relation_type: 关系类型名 roles: {角色名: 实体名} 字典 """ rel_id = f"{relation_type}_{len(self.relations)}" self.relations[rel_id] = { "type": relation_type, "roles": roles } return rel_id def query_by_role(self, relation_type, role_name, entity): """查询某实体在某角色中的所有关系""" results = [] for rel_id, data in self.relations.items(): if data["type"] == relation_type and data["roles"].get(role_name) == entity: results.append((rel_id, data)) return results def get_relation_string(self, rel_id): """人类可读的关系描述""" data = self.relations[rel_id] roles_str = ", ".join(f"{k}={v}" for k, v in data["roles"].items()) return f"{data['type']}({roles_str})" # 创建N-元关系 nary = NaryRelation(onto) r1 = nary.create("创作事件", {"作者": "鲁迅", "作品": "呐喊", "时间": "1923"}) r2 = nary.create("创作事件", {"作者": "鲁迅", "作品": "彷徨", "时间": "1926"}) r3 = nary.create("任职事件", {"人物": "鲁迅", "机构": "北京大学", "时间": "1920"}) print("=== N-元关系 ===") for rid in [r1, r2, r3]: print(f" {nary.get_relation_string(rid)}") print(" === 查询鲁迅的创作事件 ===") for rid, data in nary.query_by_role("创作事件", "作者", "鲁迅"): print(f" {nary.get_relation_string(rid)}")