Explore the intricacies of knowledge graphs and their applications in natural language processing in this 52-minute lecture from CMU's Neural Nets for NLP 2018 series. Delve into topics such as knowledge bases, structured databases, and WordNet. Examine the challenges of knowledge base incompleteness and learn about various relation extraction techniques, including Neural Tensor Networks and Hyperplane Translation. Discover the Decomposable Relation Model and methods for modeling distant supervision noise in neural models. Compare approaches for modeling word embeddings versus modeling relations, and investigate tensor decomposition. Conclude with an exploration of retrofitting embeddings to existing lexicons, gaining valuable insights into the intersection of knowledge representation and neural network technologies.
Overview
Syllabus
Knowledge Bases . Structured databases of knowledge usually containing
WordNet (Miller 1995)
Knowledge Base Incompleteness
Relation Extraction w/ Neural Tensor Networks (Socher et al. 2013)
Relation Extraction w/ Hyperplane Translation (Wang et al. 2014) • Motivation it is not realistic to assume that al dimensions are relevant to a particular relation • Solution project the word vectors on a hyperplane specifically for that relation, then verily relation
Decomposable Relation Model (Xie et al. 2017)
Modeling Distant Supervision Noise in Neural Models (Luo et al. 2017)
Modeling Word Embeddings vs. Modeling Relations
Tensor Decomposition (Sutskever et al. 2009)
Retrofitting of Embeddings to Existing Lexicons (Faruqui et al. 2015)
Taught by
Graham Neubig
Related Courses
-
Neural Nets for NLP 2021 - Neural Nets + Knowledge Bases
-
Neural Nets for NLP 2020 - Neural Nets + Knowledge Bases
-
Neural Nets for NLP 2021 - Distributional Semantics and Word Vectors
-
CMU Advanced NLP: Information Extraction and Knowledge-Based QA
-
CMU Neural Nets for NLP - Distributional Semantics and Word Vectors
-
CMU Neural Nets for NLP 2018 - Models of Words
