KDD 2014 Tutorial on Constructing and Mining Web-scale Knowledge Graphs, New York, August 24, 2014
82
Supervised relation extraction
• Sentence-level labels of relation mentions
• "Apple CEO Steve Jobs said.." => (SteveJobs, CEO, Apple) • "Steve Jobs said that Apple will.." => NIL
• Traditional relation extraction datasets
• ACE 2004 • MUC-7 • Biomedical datasets (e.g BioNLP clallenges)
• Entity classification • Group/expert detection • Ontology alignment • Object ranking
• Embedding models
KDD 2014 Tutorial on Constructing and Mining Web-scale Knowledge Graphs, New York, August 24, 2014
A sampler of research problems
• Growth: knowledge graphs are incomplete! • Link prediction: add relations • Ontology matching: connect graphs • Knowledge extraction: extract new entities and relations from web/text
•
Learn classifiers from +/- examples
•
Typical features: context words + POS, dependency path between
entities, named entity tags, token/parse-path/entity distance
知识图谱需要的技术
知识图谱架早期知识图谱架构• 知识图谱一般架构:[来源自百科]• 早期知识图谱架构
架构讨论 数据检索
规范化处 理
结果导读
预处理
可视化数 据
构建关系 矩阵网络
图谱参数 调整
知识抽取
• 实体概念抽取 • 实体概念映射 • 关系抽取 • 质量评估
•
Intelligence: can AI emerge from knowledge graphs?
• Automatic reasoning and planning
• Generalization and abstraction
KDD 2014 Tutorial on Constructing and Mining Web-scale Knowledge Graphs, New York, August 24, 2014
• Extracting semantic relations between sets of [grounded] entities
• Numerous variants:
• Undefined vs pre-determined set of relations • Binary vs n-ary relations, facet discovery • Extracting temporal information • Supervision: {fully, un, semi, distant}-supervision • Cues used: only lexical vs full linguistic features
80
Relation Extraction
? playFor
Kobe Bryant
LA Lakers
“Kobe Bryant, “Kobe
“Kobe Bryant
the franchise player of once again saved man of the match for
the Lakers” his team” Los Angeles”
• Validation: knowledge graphs are not always correct! • Entity resolution: merge duplicate entities, split wrongly merged ones
• Error detection: remove false assertions
•
Interface: how to make it easier to access knowledge?
• Semantic parsing: interpret the meaning of queries
• Question answering: compute answers using the knowledge graph
2. Entity resolution
Hale Waihona Puke • Single entity methods • Relational methods
3. Link prediction
• Rule-based methods • Probabilistic models • Factorization methods
Not in this tutorial:
9
关系抽取
定义: 常见手段:
语义模式匹配[频繁模式抽取,基于密度聚类,基于语义相似性] 层次主题模型[弱监督]
Methods and techniques
1. Relation extraction:
• Supervised models • Semi-supervised models • Distant supervision